Silicon-Micromachined Subterahertz Frequency Beam-Steered Dual-Port Array Antenna

This article presents an unbalanced-fed silicon micromachined dual-port dual-line antenna array. The radiation pattern of the antenna array can be steered in the E-plane by sweeping the frequency and can be switched between a broad and a notched beam by exciting the ports with in-phase or out-of-phase signals. The antenna is designed for and implemented by silicon micromachining. Each single-line subarray consists of #8 antenna apertures in which the field amplitude is tapered in the H-plane, and the phase imbalance of unbalanced power dividers is minimized by integrated delay sections in the feed network. The measured return loss of the antenna is better than 10 dB from 220 to 295 GHz for both input ports (29.1% fractional bandwidth). The antenna prototype is designed for 40° of beam steering in the E-plane (scanning speed of 4°/GHz) by sweeping the frequency from 238 to 248 GHz. The measured sidelobe level of the broad beam in the H-plane is better than 18.5 dB, and the measured depth of the notched beam is better than 22.5 dB in the entire scanning range. In addition to the dual-port dual-line antenna array, a single-line 1 × 8 antenna array is also implemented for reference measurement purposes. The measured return loss of the single-line antenna array is better than 10 dB from 220 to 314 GHz (35.2% fractional bandwidth), and its measured sidelobe level is between 18 and 21.3 dB in the H-plane from 220 to 280 GHz. Besides, the simulation data and the measurement results are in excellent agreement.

Abstract-This article presents an unbalanced-fed silicon micromachined dual-port dual-line antenna array.The radiation pattern of the antenna array can be steered in the E-plane by sweeping the frequency and can be switched between a broad and a notched beam by exciting the ports with in-phase or out-of-phase signals.
The antenna is designed for and implemented by silicon micromachining.Each single-line subarray consists of #8 antenna apertures in which the field amplitude is tapered in the H-plane, and the phase imbalance of unbalanced power dividers is minimized by integrated delay sections in the feed network.The measured return loss of the antenna is better than 10 dB from 220 to 295 GHz for both input ports (29.1% fractional bandwidth).The antenna prototype is designed for 40°of beam steering in the E-plane (scanning speed of 4°/GHz) by sweeping the frequency from 238 to 248 GHz.The measured sidelobe level of the broad beam in the H-plane is better than 18.5 dB, and the measured depth of the notched beam is better than 22.5 dB in the entire scanning range.In addition to the dual-port dual-line antenna array, a single-line 1 × 8 antenna array is also implemented for reference measurement purposes.The measured return loss of the single-line antenna array is better than 10 dB from 220 to 314 GHz (35.2% fractional bandwidth), and its measured sidelobe level is between 18 and 21.3 dB in the H-plane from 220 to 280 GHz.Besides, the simulation data and the measurement results are in excellent agreement.

I. INTRODUCTION
T HE millimeter-wave and subterahertz (sub-THz) fre- quency bands are getting increased attention primarily because of the need for larger bandwidth for communication, imaging, and sensing applications.Sub-THz and THz systems have various applications, including high-resolution car radars, short-range high data-rate communication, intersatellite communication, deep space exploration, and biomedical imaging.
Many of these applications demand high gain, low sidelobe level (SLL), and, most importantly, compact antennas.
Fabricating an antenna array operating at the sub-THz frequency range is challenging with conventional manufacturing techniques such as computer numerical control (CNC) machining [1], [2], and 3-D printing [3] due to poor fabrication tolerances.The required accuracy and precision for such highfrequency antennas may not be achievable with these technologies, especially for complex geometries.Furthermore, the sequential nature of these fabrication techniques increases the cost and production time, making them unsuitable for highvolume production and commercial applications.Alternatively, silicon micromachining using deep-reactive-ion-etching (DRIE) and sidewall metallization [4], [5] enables the fabrication of integrable, compact, and mass-producible devices at low cost.It offers micrometer-range fabrication tolerances and very low surface roughness down to a few nanometers, resulting in low-loss waveguides and components.The lowest waveguide loss in the 220-325 GHz frequency band is recorded in [6], 0.02-0.07dB/mm.The efficiency of the silicon micromachining techniques has already been demonstrated for different sub-THz components, such as antennas [7], [8], switches [9], phase shifters [10], and filters [11].Moreover, the authors in [12] and [13] present silicon micromachined components operating up to 2.1 and 2.7 THz, respectively, implying the huge potential of silicon micromachining techniques for developing sub-THz and THz devices.
There are many methods to taper the amplitude of the field distribution of an antenna to reduce its SLL.Horn antennas with high gain and low SLL can be designed by applying appropriate discontinuities in the opening section of the antenna wall or by loading the aperture of the horn antenna using a dielectric lens [14], [15].Due to fabrication complexities in the sub-THz frequencies, horn antennas with such small-scale features are challenging to fabricate.Besides, reflector antennas with very high gain and low SLL are widely used in space and satellite communication [16], [17] but require an extremely low surface roughness, even at lower frequencies, for wavefront uniformity on the virtual aperture of the antenna.Besides, reflector antennas are bulky, which makes it challenging to integrate them with other components.Furthermore, metallic lens antennas [18], [19] with high gain and low SLL are attractive alternatives, but they suffer from design and fabrication complexities.Alternatively, planar antenna arrays are compact and suitable for creating high-gain antennas with low SLL.The most common planar antenna arrays are based on multilayer substrate-based structures [20] and hollow waveguide-based slot arrays [21].Substrate loss plays an essential role in the efficiency of sub-THz antennas, so substrate-based arrays suffer from more loss and lower total efficiency compared to hollow waveguide arrays.As a result, utilizing silicon micromachining to implement planar antenna arrays based on hollow waveguides is preferred for systems operating in the sub-THz frequency range, especially above 200 GHz, due to their unique features, including low loss, high gain, high integrability, and high compatibility with SLL reduction techniques [22].
Active and passive beam forming are two main approaches for implementing beam-steered antenna arrays.In active beamsteering, multiple phase shifters and attenuators are needed to control the phase and amplitude of each antenna element/subarray [23].Although active beam forming is very fast and precise, implementing it in the sub-THz frequency range is extremely challenging due to immature phase shifters, complexity, cost, and high power consumption.The simplest approach for passive beam forming involves mechanically adjusting the orientation of the antenna feed to steer the beam direction [24], [25].This technique does not require complex electronic controls but relies on precise mechanical movements, which complicates its implementation in the sub-THz frequency range.
Alternatively, frequency beam steering allows for scanning the direction of the main beam without mechanically moving the antenna feed or adjusting the phase of individual elements.The principles of frequency beam steering in the sub-THz frequency have already been demonstrated in [26], [27], [28], [29], [30], and [31].Ranzani et al. [26] reported a frequency beam-steered antenna array operating at the G-band with 30°o f field of view (FOV) in 26 GHz of bandwidth.Afterward, Cullens et al. [27] presented a slot array antenna operating from 130 to 180 GHz with 32.5°of FOV and a scanning speed of 1.04°/GHz.Then, Ranzani et al. [28] elaborated on the previous designs and illustrated an improved slot array antenna operating in the 130-180 GHz frequency range with 40°of FOV and a scanning speed of 2°/GHz.Recently a (FOV, scanning-speed) of (38°, 0.42°/GHz), (45°, 0.56°/GHz), and (50°, 3.33°/GHz) are demonstrated for the antenna arrays reported in [29], [30], and [31], respectively.The trade-off between the FOV, scanning speed, appearance of the grating lobes, and spacing between antenna elements makes it challenging to design a fast scanning array antenna with a wide FOV without any grating lobe, appearing in the scanning range.
This article reports on a silicon-micromachined dual-port dual-line antenna array with fast-scanning frequency beamsteering and wide grating lobe-free FOV, in which the trade-offs between different antenna parameters are considered carefully.The designed antenna array is excited by an integrated corporate feed network, which is folded into two layers to achieve a compact footprint.The dual-line antenna array consists of two single-line 1 × 8 subarrays with separate input ports; thus, the dual-line array can be fed with two independent excitation sources.Furthermore, it utilizes different-length delay lines in each antenna branch to perform the frequency beam steering.Even though the antenna array is designed to operate properly from 220 to 300 GHz, the frequency beam steering bandwidth is designed between 238 and 248 GHz with 40°of FOV, i.e., the scanning speed of 4°/GHz, which is faster compared to the previously reported antenna arrays.Moreover, in-phase excitation results in a steered broad beam, and out-of-phase excitation leads to a steered notched beam in the E-plane of the antenna array.Besides, the SLL is reduced in the H-plane by tapering the amplitude profile of the single-line subarrays.These characteristics of the proposed antenna array make it well-suited for short-range target locating and imaging applications.

II. ANTENNA DESIGN
Fig. 1 shows a perspective view of the designed dual-port dualline antenna array, consisting of two single-line 1 × 8 subarrays, a delay network for frequency beam steering, and two E-plane transitions to standard WR-3.4 waveguide at the input ports.The antenna array can operate in the 220-295 GHz frequency band and is optimized to be fabricated by silicon micromachining, with fabrication imperfections considered and compensated in the simulation models.These fabrication imperfections include silicon underetching, chip-to-chip misalignment, lithography and mask shrinkage error, and mask misalignment, in which the silicon underetching and chip-to-chip misalignment are more dominant, i.e., they can strongly affect the RF performance of the fabricated device and are discussed with comprehensive details in [32].
Fig. 2(a) shows the designed antenna aperture with horizontal in-plane (referring to the wafer surface) waveguide feed and out-of-plane aperture opening.The size of the antenna aperture, which is used as an antenna unit cell to build the larger array, is 820 µm × 497 µm.The signal is coupled from the in-plane waveguide to the out-of-plane waveguide by a matching E-plane step, in which the width and height are 194 µm and 139 µm, respectively.Besides, a waveguide discontinuity in the vertical section is used for matching the antenna slot to the free space impedance [33], [34].The simulated realized gain of the antenna unit-cell is 8.04 dBi at 243 GHz [E-and H-plane radiation patterns are shown in Fig. 3(a)], and its simulated return loss [shown in Fig. 3(c)] is better than 10 dB from 196 to 318 GHz, 47.5% fractional bandwidth (FBW).
The power distribution (feed) network is designed using unbalanced power dividers to taper the field amplitude, thus reducing the SLL of the antenna array in the H-plane.The feed network has three dividing steps, which are folded into two layers to minimize the footprint of the antenna.As shown in Fig. 2(c), the first two dividing steps utilize T-junction power dividers with inductive posts [35], [36], and the third dividing step uses Y-junction power dividers [37].In addition, the phase imbalance imposed by the unequal dividing ratios of the unbalanced power dividers is minimized by integrated delay sections in every individual branch [22].The normalized amplitude distribution and the normalized residual phase error of each branch at 243 GHz are shown in Fig. 3(b).The amplitude profile is symmetric, with the amplitude of the corner elements being 8 dB lower than the center ones.Besides, the antenna apertures are perfectly coherent with a phase difference of less than 2.5°.The simulated return loss of the feed network is depicted in Fig. 3(c) and is better than 10 dB from 193 to 310 GHz, 46.5% FBW.
Utilizing multistep E-plane waveguide transition/bend at the input ports is a common approach, which allows for direct connection of the designed antenna array to a standard waveguide flange/interposer [38], [39].Fig. 2(b) shows a perspective view of the designed multistep E-plane waveguide transition/bend that facilitates connecting the internal in-plane waveguides to an out-of-plane standard WR-3.4 waveguide interface (864 µm × 432 µm).The designed transition includes an E-plane step inside the in-plane waveguide section, where the width and height are optimized to minimize the input reflection.The simulated performance, shown in Fig. 3(c), indicates a beyond full-band E-plane transition with a return loss of better than 13 dB from 180 to 320 GHz.
The single-line 1 × 8 subarrays are made of #8 antenna unit cells connected to the unbalanced feed network.Although reducing the distance between the antenna unit cells alleviates the grating lobe issue, it also increases the half-power-beamwidth (HPBW) of the antenna array.The optimal distance should achieve a balance between reasonable grating lobe suppression and maintaining an acceptable HPBW.According to the antenna specifications on the HPBW (8.5°-9.5°between238 and 248 GHz in the H-plane), the pitch between the antenna apertures is designed to be 980 µm in the H-plane, 0.8λ at 243 GHz, to avoid the appearance of the grating lobes in the 238-248 GHz frequency band.
The dual-line antenna array is implemented by two single-line subarrays, which are placed adjacently in the E-plane.Each single-line subarray is connected to an individual input port via a different-length delay line.The delay-line length difference leads to a linearly increasing phase difference when sweeping the frequency; thus, sweeping the frequency steers the radiation pattern of the dual-line antenna array in the E-plane.Besides, each input port can be connected to a separate frequency extender, making it possible to create different beam shapes.In-phase excitation of the ports creates a steerable broad beam, while out-of-phase excitation results in a steerable notched beam.The capability to generate both broad and notched beams makes the proposed antenna array suitable for short-range radars, target locating, and imaging applications.
According to the specifications, the designed antenna array requires a FOV of 40°in the E-plane for the 10 GHz bandwidth of 238-248 GHz, ±20°of scanning range.Even though increasing the distance between the subarrays decreases the scanning range (FOV), it improves the depth of the notch in the notched beam mode.On the other hand, reducing the distance decreases the directivity and increases the HPBW in the broad beam mode.Besides, reducing the distance decreases the required phase difference between the subarrays to achieve a particular FOV.According to these trade-offs between the FOV, HPBW, directivity, and depth of the notch, the pitch between the single-line antenna subarrays is designed to be 900 µm, 0.73λ at 243 GHz.As a result, the required phase difference for steering the beams from −20°to +20°is ±90°, leading to the delay-lines length difference of 12λ at 243 GHz.The induced phase difference in the broad/notched beam mode is −90°/90°at 238 GHz, 0°/180°a t 243 GHz, and increases to 90°/270°at 248 GHz.Therefore, the broad beam at 238/248 GHz is similar to the notched beam at 248/238 GHz.Fig. 2(d) shows a top view of the designed dual-line antenna array with detailed dimensions.A groove with a width and depth of 250 µm and 746 µm, respectively, is also added between the antenna rows to reduce the cross-coupling between the single-line subarrays.
The device is designed to be fabricated in three vertically assembled silicon-on-insulator (SOI) micromachined chips.The thickness of the utilized SOI wafer is 308 µm composed of a device layer (30 µm), a handle layer (275 µm), and a buried oxide (BOX) layer (3 µm).The handle layer defines the height of the in-plane waveguides, resulting in a slightly shallower waveguide system (858 µm × 275 µm), and the device layer forms the bottom of the in-plane waveguides.Fig. 4(a) shows the stacked chips of the dual-line antenna array and how the assembled device is mounted on a dual-channel custom-made waveguide interposer that routes the ports to two standard WR-3.4 waveguides, making it possible to feed the antenna array with two phase-locked frequency extenders.In addition to the dualport dual-line antenna array, a single-line antenna array without the delay network has also been implemented for characterizing parts of the circuit.The single-line array has one waveguide port and can be mounted directly on a standard WR-3.4 waveguide flange.Fig. 4(b) shows the stacked chips of the single-line

III. FABRICATION AND ASSEMBLY
Silicon micromachining, using DRIE and sidewall metallization, is utilized to fabricate the designed antenna array.Silicon etching is performed by an STS plasma etching tool with silicon etch rate of approximately 1 µm/min and selectivity over silicon dioxide (SiO 2 ) of 450:1.An endpoint detection system based on optical emission spectroscopy is also used in the plasma chamber to detect the exposure of the BOX layer during the deep silicon etching.
Three SiO 2 hard-masks are used to etch the silicon substrate, one for the device layer and two for the handle layer.Due to the large dimensions of the designed antenna array, the lithography and mask shrinkage errors, which are in the order of a few hundred nanometers, are negligible compared to other error sources.However, mask misalignment, especially front-to-back mask  misalignment, which is more severe than front-to-front mask misalignment, is checked under microscope and is better than ±1 µm in both x-and y-directions.The device layer hard-mask is for etching the full height of the device layer (30 µm).The first handle layer hard-mask is for etching the full-height waveguides (275 µm), and the second one is for etching the E-plane steps in the waveguides.The etching of the step has been optimized experimentally to achieve the desired etch depth of 136 µm.The etched chips are then metalized with 1.5 µm of gold on a 50 nm titanium-tungsten (TiW) adhesion layer in a dc magnetron sputtering tool.Finally, the metalized chips are bonded by thermo-compression bonding at 200 °C.Chip-to-chip alignment is done by an integrated vernier scale patterned on four corners of the chips [30].After the bonding, the chip-to-chip alignment is checked and is better than ±3 µm in both x-and y-directions, which is typical for a carefully aligned chip stack [30], [32], [40].Fig. 5(a) and (b) show two optical microscope photographs of the fabricated dual-line antenna array after gold metallization, depicting the antenna apertures and the feed network on chips #1 and #3.
Due to silicon under-etching in the waveguide trenches, the sidewalls are not completely vertical and have a slight slope.This sidewall slope can affect the RF performance of the fabricated antenna and should be considered in the simulations.Thus, 1.25°s idewall angle, predicted by test fabrication runs, is considered in the simulations to compensate for the impacts of the silicon under-etching on the RF performance.This sidewall slope assumption is a good initial approximation; however, deviation of the actual sidewall slope from this value depends strongly on additional factors, including the mask pattern, etching depth, width of the etched trench, and overall etched area on the wafer.For this reason, dummy fall-out structures are placed inside the waveguides to reduce the sidewall slope error, which the details are out of the scope of this article and are demonstrated in [41].Fig. 6 shows a cross-section scanning electron microscope (SEM) image of a fabricated trench diced with a diamond blade.As can be seen, the silicon under-etching is 6.9 µm for a 301 µm deep trench, 1.31°, which is in excellent agreement with the initial approximation.The debris visible in Fig. 6 is due to the dicing process.
Potential fabrication tolerances, quality of the silicon etching, and profile of the E-plane steps in the waveguides are also investigated by taking SEM images from different areas of the fabricated chips.Fig. 7(a) and (b) show an overview of chip #1 and chip #3 with close-up views of different elements, respectively, and Fig. 7(c) depicts a close-up view from a T-junction power divider with an inductive post.According to Fig. 7, the fabricated chips are in excellent condition and ready for thermo-compression bonding.
IV. ANTENNA CHARACTERIZATION Fig. 8 shows the CNC machined interposer, which is utilized for exciting the dual-port dual-line antenna array.The interposer, which is custom-made from brass, consists of two waveguide channels positioned 982 µm (pitch) apart from each other on one side and diverted on the other side to facilitate connecting two standard WR-3.4 waveguide flanges to the interposer.The fabricated dual-port antenna array is aligned to the interposer by internal alignment holes, in which one of them is a tightly fit circular hole, and the other is a loose elliptical hole to achieve a repeatable chip-to-interposer connection [8], [42], with an alignment accuracy of ±5 µm in x-and y-directions.
The input reflections (S 11 ) are measured using Rohde and Schwarz ZVA-24 VNA and frequency extenders (220-330 GHz).The frequency extenders are calibrated with a standard calibration kit by through-offset-short-match method to move the reference planes at the outputs of the frequency extenders.The measured return losses of the fabricated antenna arrays are compared to the simulated data and depicted in Fig. 9. Fig. 9(a) shows the measured return loss of the dual-line antenna arrays, with the S-parameters of the interposer de-embedded from the measured data by postprocessing in MATLAB, which is better than 10 dB from 220 to 295 GHz, 29.1% FBW.Besides, Fig. 9(b) shows the measured return loss of the single-line antenna array, which is better than 10 dB between 220 and 314 GHz, 35.2% FBW.As can be seen, the measured results and simulated data are in excellent agreement for both antenna array prototypes.
The radiation patterns are measured in an anechoic antenna chamber, characterized up to 750 GHz, with a 4-degree-offreedom fully automated robot, which is covered with highfrequency absorbers to reduce reflections (shown in Fig. 10).The measurement setup is configured with a standard rectangular horn antenna (with 25 dBi gain) as transmitting element, and the antenna under test (AUT), mounted on the robot, as receiving element.Moreover, the measurement setup is calibrated with another 25 dBi gain rectangular horn antenna.The radiation pattern measurement of the dual-port antenna array necessitates the utilization of two phase-locked frequency extenders in the receiving part to be able to characterize different radiation patterns.However, the single-line antenna measurement needs only one frequency extender, which simplifies the setup.Fig. 11 shows a schematic block diagram of the measurement setup for the dual-port antenna array, and Fig. 12     agreement with the simulation data.Although the simulated cross-polarization levels of both antenna arrays are better than −39.5 dBi, the measurements show values of better than −16.3 and −14.7 dBi in the 238-248 GHz and 220-300 GHz frequency ranges, respectively.This huge discrepancy is because of the cross-polarization level of the utilized standard horn antenna as the transmitting element, i.e., the whole cross-polarization measurement is limited to the cross-polarization level of the horn antenna.Furthermore, the measured SLL suppression is between 18.5 and 20.3 dB in the 238-248 GHz frequency band for the dual-line array, which is in excellent agreement with the simulation data.For the single-line array, the measured SLL suppression is between 18 and 21.3 dB from 220 to 280 GHz and deteriorates to 15 dB at 300 GHz, as expected according to the simulation data.Although the measured SLL suppression of the single-line array conforms with the simulation data, there is a slight discrepancy between them, particularly from 250 to 265 GHz, which is likely caused by fabrication errors, such as deviations in the dividing ratios of the power dividers.This deviation can strongly affect the amplitude tapering profile and ultimately deteriorate the SLL suppression.in the entire scanning range, which is a proper notched beam to be used in short-range target locating or imaging applications.The simulation data and measurement results are in excellent agreement except for a slight reduction in the scanning range from 40°to 38.5°, the broad-to-notched beam misalignment of 0.4°, and small ripples on the measured data, which is completely normal for such a high-frequency range measurement.

V. CONCLUSION
This article presented the design, fabrication, and characterization of a dual-port dual-line antenna array.The antenna array consists of two single-line 1 × 8 subarrays.Each single-line subarray is connected to an individual input port via differentlength delay lines, creating a linearly increasing phase difference between the subarray by sweeping the frequency, thus enabling frequency beam steering.Besides, each port can be excited independently, resulting in the ability to generate different beam shapes.The SLL of the subarrays is reduced in the H-plane by tapering the field amplitude profile, and the phase imbalance imposed by the unbalanced power dividers is compensated with integrated delay sections in the feed network.The antenna array generates a frequency-steered broad beam in the E-plane when the ports are excited with in-phase signals.Conversely, out-ofphase excitation signals result in a frequency-steered notched beam in the E-plane, while the notch depth remains better than 22.5 dB in the entire sweeping range.The antenna array operates properly from 220 to 295 GHz, but the designed beam steering prototype operates from 238 to 248 GHz with 38.5°of beam steering from −19.4°/−19°to 19.1°/19.5°in the broad/notched beam mode, i.e., the scanning speed is 3.85°/GHz.Furthermore, all the measurements indicated an excellent agreement between the measured results and simulated data.

Fig. 1 .
Fig. 1.Perspective view of the dual-port dual-line antenna array including the folded corporate feed networks, used for amplitude tapering in the H-plane (SLL reduction), and the delay line sections, used for frequency beam steering in the E-plane.

Fig. 2 .
Fig. 2. Perspective view of different elements of the antenna array with detailed dimensions.(a) Antenna unit cell.(b) Multistep E-plane waveguide transition.(c) Unbalanced power distribution network.(d) Top view of the dual-line antenna array with detailed dimensions.

Fig. 3 .
Fig. 3. (a) Simulated realized gain of the antenna unit cell in the E-and Hplanes at 243 GHz.(b) Normalized amplitude profile of the single-line antenna subarrays at 243 GHz along with normalized residual phase errors.(c) Simulated return loss of different elements.

Fig. 4 .
Fig. 4. Overview of the vertically stacked SOI chips for: (a) dual-line dual-port antenna array along with an illustration of the frequency beam steering and how the assembled antenna is mounted on the excitation sources, and (b) single-line 1 × 8 antenna array mounted directly on a standard WR-3.4 waveguide flange along with its radiation pattern.(c) Two perpendicular cross-section views A and B referring to Fig. 4(b).

Fig. 5 .
Fig. 5. Two optical microscope photographs of the fabricated chips of the dual-line antenna array showing the power divider sections and antenna apertures located on (a) chip #1 and (b) chip #3.

Fig. 6 .
Fig. 6.Cross-section scanning electron microscope image of a diced waveguide trench.The sidewall slope is 1.31°, indicating an excellent agreement with the initial approximation of 1.25°.

Fig. 7 .
Fig. 7. Scanning electron microscope images of the fabricated (a) chip #1 and (b) chip #3.(c) Close-up view of an inductive post inside a T-junction power divider.

Fig. 8 .
Fig. 8. Overview of the dual-channel interposer utilized for exciting the dualport antenna array with a total size of 75 mm × 20 mm × 20 mm.It is made of brass and manufactured in two pieces which are fastened together with screws.

Fig. 9 .
Fig. 9. Measured return loss compared to the simulated data for (a) dual-line and (b) single-line antenna arrays.

Fig. 10 .
Fig. 10.Configuration of the measurement setup.The AUT is mounted on the automatic robot as the receiver, sweeping both azimuth and elevation angles.
(a) and (b) show the antennas mounted on standard waveguide flanges in the setup.Fig. 13(a) and (b) show the simulated and measured normalized H-plane radiation patterns for the dual-line antenna array from 238 to 248 GHz and the single-line antenna array

Fig.
Fig. Schematic block diagram of the measurement setup used for characterizing the radiation patterns of the dual-port dual-line antenna array.

Fig. 12 .
Fig. 12. Configuration of the measurement setup with the mounted (a) dualport dual-line antenna array on the custom-made interposer and (b) single-line antenna array on a standard WR-3.4 waveguide flange.

Fig. 13 .
Fig. 13.Normalized H-plane radiation patterns of (a) beam-steering dual-line antenna array in the scanning range of 238-248 GHz and (b) fixed-beam singleline antenna array from 220 to 300 GHz, measured in an anechoic antenna chamber and compared with the simulation data.

Fig. 14 .
Fig. 14.Simulated and measured maximum realized gain for both co-and cross-polarizations and the SLL in the H-plane versus frequency for (a) dual-line antenna array in the scanning range of 238-248 GHz and (b) single-line antenna array from 220 to 300 GHz.

Fig. 15 .
Fig. 15.Simulated and measured E-plane radiation patterns of the dual-port dual-line antenna array at different frequencies with (a) in-phase and (b) out-ofphase excitation signals.

Fig. 15
shows the measured and simulated E-plane radiation patterns of the dual-line antenna array at different frequencies when the ports are excited with in-phase [Fig.15(a)] and out-of-phase [Fig.15(b)] signals.In the in-phase state, both ports are excited with the same phase from 238 to 248 GHz, leading to a steered broad-beam pattern, and in the out-of-phase state, the ports are excited with out-of-phase signals, leading to a steered notched-beam pattern in the E-plane.While the position of the maximum in the broad beam is at −19.4°and 19.1°, the notch is positioned at −19°and 19.5°in the notched beam at 238 and 248 GHz, respectively, indicating an excellent broad-to-notched-beam alignment.Besides, the profile of the broad beam at 238/248 GHz is similar to that of the notched beam at 248/238 GHz, which is expected according to the simulation data.The depth of the notched beam remains better than 22.5 dB Silicon-Micromachined Subterahertz Frequency Beam-Steered Dual-Port Array Antenna Armin Karimi , Graduate Student Member, IEEE, Umer Shah , Senior Member, IEEE, Alireza Madannejad , Student Member, IEEE, and Joachim Oberhammer , Fellow, IEEE