Broad band on-chip Ti-based RF terminator for high-speed photonic platforms

Abstract

In this paper, we demonstrate a compact broadband on-chip radio-frequency (RF) termination based on pure titanium thin-film resistors integrated into coplanar waveguide (CPW) structures. The terminated transmission lines exhibit good impedance matching, with simulated reflection coefficients below −20 dB over 100 GHz bandwidth and experimentally measured reflections below −10 dB up to 65 GHz. Additional measurements evaluate the stability and power-handling behavior of the Ti resistors. The results show that thin-film titanium can operate as an effective on-chip RF termination in integrated CPW structures, maintaining stable RF performance over repeated measurements. Such terminations are relevant for high-speed photonic platforms where compact integration and reduced packaging complexity are desirable, including dense modulator architectures.

Graphical abstract

Keywords

Traveling-wave electrodes (TWEs) / Broadband impedance matching / DWDM systems / Coplanar waveguide (CPW) / On-chip radio-frequency (RF) termination

1 Introduction

The increasing demand for high-speed, broadband communication systems has driven the development of integrated microwave and photonic components capable of operating efficiently at frequencies exceeding tens of gigahertz. Among these, coplanar waveguide (CPW) structures have emerged as a preferred transmission line geometry due to their planar nature, ease of integration with optoelectronic devices, and compatibility with standard Silicon fabrication processes.

In high-speed, radio-frequency (RF) and microwave systems, impedance matching is essential to ensure efficient signal transmission and suppress reflections that degrade performance. A 50 Ω termination is commonly used to match the characteristic impedance of transmission lines such as coaxial cables or coplanar waveguides (CPWs). Without proper termination, reflected signals interfere with the incident wave, causing standing waves, signal distortion, and loss of modulation fidelity. These effects become increasingly critical at frequencies above 10 GHz [1]. This principle is especially important in electro-optic modulators, which often employ traveling wave electrodes (TWEs) to apply RF signals across the modulating region. To preserve signal integrity over a wide bandwidth, the RF line must be terminated with a matched load at the end of the electrode. Traditionally, this is achieved using external 50 Ω terminators, connected via RF probes or wire bonds. However, as photonic integration scales, particularly in wavelength division multiplexing (WDM) systems [2], IQ modulators, dual-polarization modulators [3], or modulator array [4], external termination becomes increasingly impractical.

To address the challenges of scaling broadband RF terminations in high-speed photonic systems, particularly in dense modulator arrays for DWDM applications, on-chip RF termination resistors have become an attractive approach. These terminations are integrated directly at the end of traveling-wave electrodes and are commonly implemented using thin-film metals such as nickel–chromium (NiCr) [5] or titanium nitride and related alloys (TiN) [6,7]. Both materials are well established in silicon photonics and CMOS-compatible platforms due to their relatively low temperature coefficients of resistance and documented long-term stability under controlled fabrication process.

Pure titanium on the other hand, is more widely employed in integrated photonics as a DC resistive element, for example in microheaters and thermo-optic tuning structures [8–11] and has been less explored as a broadband RF termination material with detailed scattering-parameter characterization. A practical motivation for considering Ti in this context is the potential to co-integrate RF terminations and DC resistive elements within a single-metal lithography and lift-off process, enabling both functions to be implemented using the same material system and fabrication sequence.

From a process point of view, Ti thin films are simple to integrate. They can be deposited by standard e-beam evaporation or sputtering and patterned by lift-off, without the need for reactive gases or tight control of alloy composition, and its resistivity is predominantly dependent on deposition rate and ambient oxygen [12]. This motivates investigating whether pure Ti can also be used effectively as an RF termination resistor, enabling high-speed terminations and thermo-optic heaters to share the same simple process module - a prospect as advanced photonic integrated circuits require both high-speed drive and thermal biasing at the same time.

A central materials consideration is that titanium spontaneously forms oxide when exposed to air. Recent surface and materials studies show that this native oxide is relatively thin and predominantly TiO₂. XRD and XPS measurements on pure Ti indicate that air exposure produces a native titanium oxide layer with thickness below ~10 nm, while the metallic Ti substrate still dominates [13–15]. Together, these observations support the commonly adopted picture of a rapidly formed but self-limiting TiO₂-based film of only several nanometers thickness under ambient laboratory conditions, with the underlying Ti remaining metallic. The further growth of such oxide films over longer periods is usually slow. Classical metal oxidation theory and more recent unified models show that as the oxide becomes thicker, the effective parabolic rate constantly decreases, and the growth rate drops with time [14,16].

Within this materials context, the aim of the present work is to evaluate the RF and microwave behavior of thin-film Ti used as an on-chip RF termination resistor in high-speed photonic platforms. We investigate broadband RF termination using pure Ti thin-film resistors integrated into realistic coplanar-waveguide (CPW) transmission line environment. Full-wave electromagnetic simulations (ANSYS HFSS) are used to optimize the CPW geometry and quantify conductor and substrate losses. The designs are implemented on high-resistivity silicon wafers using a two-step lithography and lift-off process. The devices are air cladded with no passivation layer. Experimental characterization with a vector network analyzer (VNA) shows good agreement with simulation, with low reflection (S₁₁ < −20 dB) and moderate transmission loss (~5.5 dB/cm) over a 100 GHz simulated bandwidth and up to 65 GHz in measurement. The electrical and RF characteristics remain stable across repeated measurements performed about two weeks after fabrication and again after four months of storage under ambient laboratory conditions. Under continuous small-signal RF excitation of over 100 GHz and DC power dissipation of roughly 25–70 mW for more than one hour, we do not observe a significant drift in the reflection response. A further detailed study of extended long-term accelerated aging, and lifetime reliability under extended electrical or thermal stress remain interesting topics for future investigation, though these results indicate that native-oxide formation on thin-film Ti does not prevent stable high-speed RF termination over the time scales examined here.

2 Design and simulation

The design and simulation of the coplanar waveguide (CPW) transmission line with integrated titanium (Ti) thin-film termination resistors were carried out using ANSYS HFSS, a full-wave electromagnetic solver for high-frequency structures. The objective was to realize a broadband termination matched to the characteristic impedance of the CPW.

The CPW is modeled as a two-port structure with wave ports, and the simulation domain is enclosed by radiation boundaries to emulate an open environment. Material properties are assigned using standard conductivity values for gold (σ ≈ 4.1 × 10⁷ S/m), while the high-resistivity silicon substrate is treated as lossless over the simulated frequency range. A frequency sweep up to 120 GHz is performed to extract the S-parameters and characteristic impedance of the line. The simulated cross-section and representative electric field magnitude at 60 GHz are shown in Fig. 1a, and the corresponding characteristic impedance and S-parameters of the reference two-port CPW structure are shown in Fig. 1b. In this simulation both wave ports are defined with a 50 Ω port impedance to represent a matched measurement environment rather than a physical termination. This configuration represents a baseline transmissionline model used to evaluate the CPW impedance and transmission behavior prior to introducing the on-chip termination. The simulated transmission line length is 10 mm, and the resulting S₂₁ corresponds to approximately −5.5 dB transmission over this length, which is equivalent to a propagation loss of about 5.5 dB/cm.

To implement the on-chip termination, the second port is replaced by a Ti thin-film resistor bridging the signal and ground electrodes, as illustrated in the three-dimensional view in Fig. 1c. Titanium is modeled using an effective conductivity consistent with the fabrication process (σ ≈ 5.5 × 10⁵ S/m). The resistor geometry is defined by its length (14 µm), width (w), and thickness (100 nm), and the nominal resistance is estimated using the sheet-resistance relation:

R=\frac{\rho \cdot L}{A}=\frac{1}{\sigma }\cdot \frac{L}{w\cdot t}.

The ground electrode is smoothly tapered toward the signal electrode to minimize discontinuities and associated parasitic reflections. The integrated structure is then simulated to evaluate the reflection coefficient and broadband impedance matching. Figure 1d shows the simulated S₁₁ for different resistor widths, indicating increased reflection for geometries deviating from the nominal design due to impedance mismatch at the termination.

3 Fabrication and characterization results

3.1 Fabrication

The devices were fabricated on 4-inch high-resistivity silicon wafers (> 10 kΩ·cm) using standard optical lithography and lift-off processes. The layout includes two-port CPW transmission lines with lengths ranging from 300 µm to 10 mm and one-port terminated structures incorporating integrated Ti resistors. The Ti resistor width was varied from 2 µm to 10 µm to account for fabrication tolerances and to study the effect of geometry on termination performance.

A 3 µm-thick SiO₂ layer was deposited by plasma-enhanced chemical vapor deposition (PECVD) to electrically isolate the metal structures from the silicon substrate and reduce RF substrate losses. The Ti resistors were defined in the first lithography step using a 1.5 µm-thick image-reversal photoresist, followed by electron-beam evaporation of a 100 nm Ti film and lift-off. The second lithography step defined the CPW signal and ground electrodes, followed by deposition of a Ti/Au metal stack (5 nm/600 nm) and lift-off. At the termination region, the CPW signal electrode is smoothly tapered to match the width of the Ti resistor, providing a gradual impedance transition and minimizing discontinuities. An overview of the processed wafer, representative two-port CPW structures, and the one-port terminated transmission lines with integrated Ti resistors are shown in Fig. 2a–d.

3.2 Characterization

The fabricated CPW transmission lines and titanium-terminated structures were characterized using an Anritsu 37397D Lightning Vector Network Analyzer (VNA) with 100-µm pitch ground-signal-ground (GSG) probes from GGB Industries. The measurement setup was calibrated to the probe tips using a CS5 calibration substrate also from GGB Industries. Both two-port and one-port configurations were evaluated to extract scattering parameters and validate the simulation results.

To evaluate the performance of the fabricated CPW transmission lines, S-parameters were measured up to 65 GHz. Devices with lengths ranging from 50 µm to 10 mm were probed at both ends using the GSG probes. The shortest line (50 µm) served as a reference to verify calibration of system-level losses such as probe contact resistance and cable attenuation. As shown in Fig. 3, the 50 µm line exhibited negligible insertion loss after calibration, confirming the effectiveness of the measurement setup and calibration.

As the line length increased, the transmission coefficient (S₂₁) gradually decreased, reaching approximately −5.5 dB at 65 GHz for the 10 mm long CPW. This trend is consistent with simulation results and corresponds to a propagation loss of ~5.5 dB/cm at 65 GHz as mentioned in simulation results in Fig. 1b, primarily due to conductor losses in the gold electrodes and minor substrate effects. The reflection coefficient (S₁₁) was slightly higher than predicted by simulation, with peaks reaching around −10 dB. This deviation is attributed to fabrication tolerance, specifically, a reduction in electrode width and an increase in the signal-to-ground gap caused by lithographic and lift-off limitations. While these variations impact impedance matching, the results were consistent across the wafer, indicating good process uniformity. It is important to note that the reflection behavior observed in this two-port configuration with the second port connected to an external calibrated 50 Ω load, serves as a baseline for comparison with the one-port structures employing on-chip titanium resistors. To aid the comparison, transmission line parameters, Z₀ (characteristic impedance) and γ = α + jβ (complex propagation constant) were extracted from the measured S-parameters. The extraction is based on the L-2L method reported in Ref. [17] using the transmission line lengths of 500-µm for L and 1-mm for 2L.

As shown in Fig. 4a, the extracted characteristic impedance of the CPW line is approximately 67 Ω, which explains the elevated reflection levels. This impedance mismatch highlights the sensitivity of the system to small geometric deviations and underscores the importance of precise fabrication for achieving optimal termination. Some dispersion in the extracted characteristic impedance is observed above approximately 40 GHz. This could simply be an artifact stemming from the assumption of a shunt capacitance, series inductance second-order lumped model for the pad structures assumed during the L-2L extraction method (see Ref. [17] for details). Imaginary components of the characteristic impedance is also reported. The extracted imaginary part remains small compared to the real component across the measured bandwidth, indicating that the line behavior is well described by a low-loss CPW approximation and that the observed reflection levels are primarily governed by the real impedance mismatch rather than by reactive loading at the termination.

The effective dielectric constant, {\epsilon }_{\text{eff}}, is extracted from the imaginary part of the complex propagation constant, γ, and plotted in Fig. 4b. The effective dielectric constant remains relatively constant at approximately 3.5 across the measured frequency range, indicating that the longest line used for the L-2L extraction corresponds to approximately 0.4λ at 65 GHz. The attenuation constant, α, is extracted from the real part of the complex propagation constant, γ, and is also plotted in Fig. 4b. The extracted attenuation constant (in dB/mm) exhibits a \sqrt{f} dependence, where f is the frequency. This is consistent with the previous statement that conductor losses in the gold electrodes are the main loss mechanism in the fabricated transmission lines.

To evaluate the performance of on-chip RF termination, one-port CPW structures incorporating titanium thin-film resistors were characterized by measuring the reflection coefficient (S₁₁) over a frequency range from 0.5 GHz to 65 GHz. In this configuration, only one port of the VNA is connected to the input of transmission line using a GSG probe. Devices with Ti resistor widths ranging from 3 µm to 10 µm were fabricated and tested to assess impedance matching.

As shown in Fig. 5a, the 3 µm wide Ti resistor exhibited the lowest reflection across the measured frequency band, with S₁₁ remaining below −15 dB from 0.5 GHz to 65 GHz. This indicates good impedance matching to the 50 Ω system reference. The optimal width differed slightly from simulation predictions, which suggested a 5 µm width for ideal matching. This discrepancy is attributed to a lower-than-expected resistivity of the deposited Ti film, likely due to reduced oxidation during e-beam evaporation. Such variations are common and highlight the importance of process control in achieving precise impedance targets.

To further validate the effectiveness of the 3 µm wide Ti resistor, we fabricated a series of one-port CPW structures with varying transmission line lengths, ranging from 300 µm to 10 mm. As shown in Fig. 5b, all devices demonstrated reflection coefficients below −10 dB across the measured frequency range, confirming that the on-chip Ti termination performs comparably to an external 50 Ω load. This result underscores the scalability and robustness of the integrated termination approach, even for long transmission lines.

Notably, the reflection curves for longer lines exhibit periodic peaks and dips, a behavior characteristic of standing wave formation in distributed transmission lines. These oscillations arise when the physical length of the line approaches or exceeds half the guided wavelength (λ/2), leading to constructive and destructive interference (separated by an integer multiple of λ/4) between incident and reflected waves. So, the periodicity of these features is governed by the effective wavelength in the CPW structure [1]. Based on simulation results and the observed trend of increasing loss with frequency, it is reasonable to expect that the Ti-terminated structures will maintain effective impedance matching even at higher frequencies. The smooth frequency response and strong agreement with simulation suggest that the termination behavior remains stable over 100 GHz regime, making this approach suitable for next-generation high-speed photonic systems.

To examine the effect of natural aging on the Ti thin-film terminations, the same devices were re-measured approximately four months after fabrication, following the initial characterization performed two weeks post-processing, while stored under ambient laboratory conditions. The extracted DC resistance and broadband RF response remained unchanged within measurement uncertainty, indicating that the formation of a native surface oxide did not produce a measurable impact on the termination performance over this timescale (Fig. 6a).

The current-handling capability of the Ti resistors was evaluated using DC I–V measurements with the current compliance set to 100 mA and the voltage swept from 0 to 3 V with a settling interval at each step to ensure thermal stabilization. For 100 nm-thick Ti resistors with a fixed length of 14 µm and widths ranging from 3 µm to 10 µm, failure occurred at dissipated powers between approximately 25 mW and 70 mW, with wider resistors exhibiting higher power tolerance (Fig. 6b). This trend is consistent with reduced current density and improved heat spreading in wider films. Since the resistance scales with the ratio of length to width for a given sheet resistance, the geometry can be adjusted to maintain a target resistance while increasing the cross-sectional area to accommodate higher power dissipation.

In RF practice, termination loads are commonly specified by power rating. For typical modulator or CMOS driver levels, a few volts peak-to-peak into a 50 Ω load corresponds to power dissipation in the order of tens of milliwatts (e.g., 3 V_pp ≈ 1.06 V_rms, corresponding to ~22 mW in 50 Ω). The measured DC failure range therefore suggests that, within the scope of typical RF drive conditions encountered in high-speed photonic modulators, the demonstrated Ti terminations operate below their current-density-limited regime. For traveling-wave Mach–Zehnder modulators, a significant fraction of the RF power is also dissipated along the transmission line itself due to conductor and dielectric losses, further reducing the power delivered to the terminal load.

To assess stability under combined electrical stress, a 3 µm-wide Ti termination was operated under continuous DC bias corresponding to approximately 20 mW dissipation while simultaneously driven with a small-signal RF excitation up to 110 GHz using a bias-tee for more than one hour. The RF stimulus was provided by the VNA with an output power of approximately −17 dBm, corresponding to a small-signal excitation level that does not significantly contribute to device heating. The S₁₁ response of a 1 cm-long-terminated transmission line was monitored throughout this period and showed no significant drift in impedance or reflected power (Fig. 6c). For comparison, a resistor intentionally driven beyond its failure point exhibited a pronounced increase in reflection across the measured bandwidth, consistent with the loss of effective termination behavior (Fig. 6d). For clarity, a moving-average filter was applied to these measured data to reduce high-frequency measurement noise and better visualize trends in the response.

4 Discussion: broadband matching and packaging in DWDM PICs

Dense WDM transmitters concentrate many traveling-wave modulators and drivers on a single die or co-packaged module, which pushes RF interfaces to their limits. In practice, the electrical fan-out around the chip perimeter, the bond-pad real estate, and the RF path from the driver to each modulator all become system bottlenecks that cap usable bandwidth and induce inter-channel coupling. Recent multi-channel demonstration acknowledges these packaging-driven limits and the need for architecture that minimizes the electrical path and is well terminated as channel counts scale [18]. A central culprit is the bond-wire interconnect. In a 16-channel co-designed silicon transmitter, the authors compare a lane operating alone with neighboring lanes transmitting concurrently and report only a ~0.1 dB change in optical SNR; they conclude that the remaining crosstalk, reflections, and bandwidth degradation are primarily caused by bond-wire inductance, and note that shortening or eliminating the bond wires mitigates these effects [18].

Wire-bonds in the RF path act like small inductors in series. Even short, millimeter-scale bonds add enough reactance to upset a 50 Ω match, which detunes the load and shifts resonances. In an experimental study, ignoring the wire-bond led to about 6 dB worse input return loss, a noticeable resonance shift, roughly 5 dB less output power, and a large efficiency drop, showing that the bond can dominate the interface if left in the signal path [19]. At the module level, even carefully designed CPW feed networks that must taper down to wire-bond pads show bandwidth ceilings and resonance features, once bond-wires are included in the RF chain, illustrating how the edge-bonded interface itself becomes a throughput limiter as channels multiply [20]. Termination quality at the modulator load is equally consequential for system-level bandwidth and flatness. Compact electro-optic models validated against measurement show that matching the load impedance to the traveling-wave line roughly doubles the 3 dB EO bandwidth, while source matching alone yields a smaller improvement; the load side dominates because it governs reflection build-up along the electrode [21].

Interfacing 50 Ω drivers with traveling-wave modulators whose electrode impedance is lower than 50 Ω, requires an impedance-transforming interconnect and a termination matched to the line’s characteristic impedance. Co-designed tapers and matching networks that bring the modulator port toward its target impedance and present a properly matched load can flatten the electro-optic transfer and raise the electro-optic (E/O) 3 dB bandwidth (reported up to ~63 GHz), but they add circuit/layout complexity and remain sensitive to small packaging parasitics from pads, wire-bonds, vias, and process spread; modest deviations from the intended impedance can erode the achieved bandwidth and passband flatness [22]. Incorporating the termination at the device-plane reduces the transformation burden on the interconnect by removing series bond-wire inductance at the load.

These observations motivate the termination physically onto the photonic die, close to each modulator port. On-chip terminations in InP platforms have been designed, measured, and incorporated into co-design models. With appropriate resistance and sufficient shunt capacitance they reproduce the broadband behavior of an ideal load over the band of interest while avoiding the extra inductive loop area and stubs inherent to off-chip returns [22]. In direct comparison, introducing 1 mm wire-bonds at both the source and the termination reduces the modulator’s 3 dB EO bandwidth from 30 GHz to 14 GHz for a 50 Ω interface, and from 65 GHz to 18 GHz when the source and termination are matched to the ~22 Ω electrode, underscoring the penalty of off-chip connections [22]. Together, this body of evidence supports an architectural direction in which each channel’s RF path is terminated on-chip and return currents locally, reducing bond-wire inductance in the load path, suppressing reflections that accumulate along traveling-wave electrodes, and easing layout pressure on the chip perimeter as channel counts increase in DWDM systems. Within that context, the titanium thin-film CPW termination demonstrated here is a step toward DWDM-scale arrays: it supplies a compact, broadband, well-matched load at the point of use, simplifying packaging, reducing sensitivity to driver-modulator impedance mismatches.

5 Conclusion

This work investigated the use of pure titanium thin-film resistors as on-chip RF terminations integrated into coplanar waveguide (CPW) transmission lines for high-speed photonic platforms. Full-wave simulations and experimental measurements were used to evaluate the termination behavior and transmissionline performance. The fabricated CPW structures exhibited propagation losses of approximately 5.5 dB/cm at 65 GHz, and the integrated Ti resistors provided effective termination within the measured bandwidth, with reflection coefficients below −10 dB up to 65 GHz. Stability tests performed over repeated measurements and under combined RF–DC stress showed no measurable change in the RF response within the examined time scale. Within the scope of the presented fabrication and measurement framework, these results indicate that thin-film titanium can function as an on-chip RF termination in integrated CPW structures. Such termination is relevant for photonic platforms where compact integration, simplified fabrication and reduced packaging complexity are desirable, including high-speed modulator architectures.

References

[1] Pozar , D.M. : Microwave Engineering, 4th ed. Wiley, Hoboken, NJ, USA (2012)

[2] Wang , Y. , Wang , S. , Parsons , R. , Novick , A. , Gopal , V. , Jang , K. , Rizzo , A. , Chiu , C.P. , Hosseini , K. , Hoang , T.T. , Shumarayev , S. , Bergman , K. : Silicon photonics chip I/O for ultra-high-bandwidth and energy-efficient die-to-die connectivity. In: Proceedings of the IEEE Custom Integrated Circuits Conference (CICC), Denver, CO, USA, pp. 1–8 (2024)

[3] Wang , X. , Shang , C. , Pan , A. , Cheng , X. , Gui , T. , Yuan , S. , Gui , C. , Zheng , K. , Zhang , P. , Song , X. , Li , Y. , Li , L. , Zeng , C. , Xia , J. : Thin-film lithium niobate dual-polarization IQ modulator on a silicon substrate for single-carrier 1.6 Tb/s transmission. APL Photonics 7(7), 076101(2022)

[4] Li , C. , Zhang , H. , Zhou , G. , Lu , L. , Jin , M. , Dong , J. , Zhou , L. , Chen , J. : Hybrid WDM-MDM transmitter with an integrated Si modulator array and a micro-resonator comb source. Opt. Express 29(24), 39847–39858(2021)

[5] Sun , S. , He , M. , Xu , M. , Zhang , X. , Ruan , Z. , Zhou , L. , Liu , L. , Liu , L. , Yu , S. , Cai , X. : High-speed modulator with integrated termination resistor based on hybrid silicon and lithium niobate platform. J. Lightwave Technol 39(4), 1108–1115(2021)

[6] Malmros , A. , Südow , M. , Andersson , K. , Rorsman , N. : TiN thin film resistors for monolithic microwave integrated circuits. J. Vac. Sci. Technol. B Nanotechnol. Microelectron 28(5), 912–915(2010)

[7] Shanidze , L.V. , Tarasov , A.S. , Rautskiy , M.V. , Zelenov , F.V. , Konovalov , S.O. , Nemtsev , I.V. , Voloshin , A.S. , Tarasov , I.A. , Baron , F.A. , Volkov , N.V. : Cu-doped TiN_xO_y thin film resistors: DC/RF performance and reliability. Appl. Sci. (Basel) 11(16), 7498(2021)

[8] Parra , J. , Navarro-Arenas , J. , Sanchis , P. : Silicon thermo-optic phase shifters: a review of configurations and optimization strategies. Advanced Photonics Nexus 3(4), 044001(2024)

[9] Wei , Y. , Cheng , J. , Wang , Y. , Zhou , H. , Dong , J. , Huang , D. , Li , F. , Li , M. , Wai , P.K.A. , Zhang , X. : Strategy for low-loss optical devices when using high-loss materials. Adv. Photon. Res 3(9), 2200120(2022)

[10] Wang , J. , Paesani , S. , Ding , Y. , Santagati , R. , Skrzypczyk , P. , Salavrakos , A. , Tura , J. , Augusiak , R. , Mančinska , L. , Bacco , D. , Bonneau , D. , Silverstone , J.W. , Gong , Q. , Acín , A. , Rottwitt , K. , Oxenløwe , L.K. , O’Brien , J.L. , Laing , A. , Thompson , M.G. : Multidimensional quantum entanglement with large-scale integrated optics. Science 360(6386), 285–291(2018)

[11] Paesani , S. , Ding , Y. , Santagati , R. , Chakhmakhchyan , L. , Vigliar , C. , Rottwitt , K. , Oxenløwe , L.K. , Wang , J. , Thompson , M.G. , Laing , A. : Generation and sampling of quantum states of light in a silicon chip. Nat. Phys 15(9), 925–929(2019)

[12] Devulapalli , V. , Bishara , H. , Ghidelli , M. , Dehm , G. , Liebscher , C.H. : Influence of substrates and e-beam evaporation parameters on the microstructure of nanocrystalline and epitaxially grown Ti thin films. Appl. Surf. Sci 562, 150194(2021)

[13] McCafferty , E. , Wightman , J.P. : An X-ray photoelectron spectroscopy sputter profile study of the native air-formed oxide film on titanium. Appl. Surf. Sci 143(1), 92–100(1999)

[14] Xu , Z. , Rosso , K.M. , Bruemmer , S.M. : Metal oxidation kinetics and the transition from thin to thick films. Phys. Chem. Chem. Phys 14(42), 14534–14539(2012)

[15] Singh , S. , Jejusaria , A. , Singh , J. , Vashishath , M. , Kumar , D. : Comparative study of titanium, platinum, and titanium nitride thin films for micro-electro mechanical systems (MEMS) based micro-heaters. AIP Adv 12(9), 095202(2022)

[16] Peraldi , R. , Monceau , D. , Pieraggi , B. : Correlations between growth kinetics and microstructure for scales formed by high-temperature oxidation of pure nickel. II. Growth kinetics. Oxid. Met 58(3-4), 275–295(2002)

[17] Johansen , T.K. , Jiang , C. , Hadziabdic , D. , Krozer , V.: EM simulation accuracy enhancement for broadband modeling of on-wafer passive components. In: Proceedings of the European Microwave Integrated Circuit Conference (EuMIC 2007), Munich, Germany, pp. 447–450 (2007)

[18] Shi , J. , Jin , M. , Yang , T. , Shu , H. , Yang , F. , Liu , H. , Tao , Y. , Deng , J. , Chen , R. , Han , C. , Qi , N. , Wang , X. : 16-channel photonic–electric co-designed silicon transmitter with ultra-low power consumption. Photon. Res 11(2), 143–152(2023)

[19] Cavalcanti , H.M. , Manera , L.T. : A simulation methodology for wirebonds interconnects of radiofrequency integrated circuits. Journal of Integrated Circuits and Systems 10(3), 181–186(2015)

[20] Yao , W. : Towards a High-Capacity Multi-Channel Transmitter in Generic Photonic Integration Technology. PhD Thesis, Eindhoven University of Technology, Eindhoven, The Netherlands (2017)

[21] Meighan , A. , Yao , W. , Wale , M. , Williams , K.A. : Microwave model for optimizing electro-optical modulation response of the Mach–Zehnder modulator. In: Proceedings of the 22nd Annual Symposium of the IEEE/LEOS Benelux Chapter, Eindhoven University of Technology, Eindhoven, The Netherlands (2017)

[22] Meighan , A. : Co-design of the High-Speed Photonic and Electronic Integrated Circuits. PhD thesis, Eindhoven University of Technology, Eindhoven, The Netherlands (2021)

RIGHTS & PERMISSIONS

The author(s)