High-Performance Silicon Nanowire VHF Nanoelectromechanical Resonators
6.3 Metallized Si NWs as VHF Resonators
VHF resonators are first realized with metallized Si NWs. Because the Si NWs typically have intrinsic resistances in the ~1−10kΩ range and even in the ~10−100kΩ range, this poses a challenge to radio-frequency (RF), especially VHF (strictly 30−300MHz) resonance detection due to the large impedance mismatch with RF electronics (with 50Ω standard), therefore we metallize these devices for better impedance matching with RF measurement components. Similar to typical VHF top-down nanomechanical resonators [25], metallization layers consisting of 5nm Ti atop 30nm Al have been deposited onto the Si NWs by either thermal or e-beam evaporation. The samples are slightly tilted during metal deposition so that the inner walls of a microtrench are not continuously coated, while allowing the Si NW to be conformally metallized (as long as it is not very deep down into the microtrench), thus the Si NW remains the only electrical path bridging the microtrench. Further probing characterizations are performed to verify that after metallization any two pads are electrically open unless bridged by one or more Si NWs. Metallized Si NWs usually have resistances of about 70−120Ω at room temperature and are very close to 50Ω at low temperatures. We employ the bridge circuit readout scheme [26] incorporating pairs of Si NWs to be able to directly start with
devices of f0∼200MHz in the more attractive VHF/UHF ranges. Although single-device based one-port reflection detection could be used for up to 100~200MHz devices [26,27], the two-port bridge detection scheme is proven to be better, especially for ≥100MHz devices [25,26].
Fig. 6.2 demonstrates the detected signals for a typical pair of metallized Si NWs. One of the device has dimensions L=2.1μm, d=118nm (aspect ratio≈18), with detected resonance frequency ~188MHz and Q≈2500; the other device has L=2.25μm, d=142nm (aspect ratio≈16), with resonance frequency ~200MHz and Q≈2000. As shown in Fig.
6.2, for the fixed magnetic field (B) bias condition, as the RF drive power is increased, the resonance response amplitude increases and the response approaches the nonlinear regime. At the fixed RF drive condition, the resonance response increases with enhanced B field, with the voltage signal amplitude having a B2 dependency, which is a confirmed characteristic of the magnetomotive transduction [25]. In these measurements we have achieved a dramatically large signal readout with a very high signal-to-background ratio of up to 12dB as shown in Fig. 6.2 (a) and (c), by employing the high-resolution bridge- balancing and background-nulling techniques [28].
Another observation is that the Q of the 188MHz device is higher than the 200MHz device, which is consistent with the well-known f0 versus Q trade-off [25]. Since the two devices have roughly the same aspect ratio, they may have clamping loss to roughly the same extent [29]. The lower Q of the 200MHz device may also be correlated to the influence of its backward growth (as shown in the inset of Fig. 6.2 (d)), which can be effectively viewed as a free-standing Si NW cantilever device sharing one anchoring point with the 200MHz Si NW, and thus may introduce extra mechanical dissipation.
187.6 187.8 188.0 188.2 -68
-66 -64 -62 -60 -58
Transmission (dB)
Frequency (MHz)
RF Drive to the Device:
-61dBm to -41dBm, with 1dB step
(a) 187.6 187.8 188.0 188.2 -68
-66 -64 -62 -60 -58
Transmission (dB)
Frequency (MHz)
RF Drive to the Device:
-61dBm to -41dBm, with 1dB step
(a) 187.0 187.5 188.0 188.5
0 4 8 12
Transmission (dB)
Frequency (MHz)
B Field Sweep:
B=0T to 8T with 1T Step
12dB
(b)
187.0 187.5 188.0 188.5 0
4 8 12
Transmission (dB)
Frequency (MHz)
B Field Sweep:
B=0T to 8T with 1T Step
12dB
187.0 187.5 188.0 188.5 0
4 8 12
Transmission (dB)
Frequency (MHz)
B Field Sweep:
B=0T to 8T with 1T Step
12dB
(b)
199.4 199.6 199.8 200.0 -70
-68 -66 -64 -62
Transmission (dB)
Frequency (MHz)
RF Drive to the Device:
-50.5dBm to -35.5dBm with 1dB Step
(c) 199.4 199.6 199.8 200.0 -70
-68 -66 -64 -62
Transmission (dB)
Frequency (MHz)
RF Drive to the Device:
-50.5dBm to -35.5dBm with 1dB Step
(c) 199.0 199.5 200.0 200.5
0 2 4 6 8
Transmission (dB)
Frequency (MHz)
B Field Sweep B=0T to 8T with 1T Step
(d)
199.0 199.5 200.0 200.5 0
2 4 6 8
Transmission (dB)
Frequency (MHz)
B Field Sweep B=0T to 8T with 1T Step
199.0 199.5 200.0 200.5 0
2 4 6 8
Transmission (dB)
Frequency (MHz)
B Field Sweep B=0T to 8T with 1T Step
(d)
Fig. 6.2 Metallized Si NWs as VHF NEMS resonators. (a) and (b) are for the 187.8MHz device with drive RF power sweep and magnetic field sweep, respectively, with the magnetomotive transduction and a bridge detection scheme. Inset in (b) is the SEM image of the device. (c) and (d) are the drive RF power sweep and the magnetic field sweep, respectively, for the 199.7MHz device. Inset in (d) is the device SEM image where the backward Si NWs growth is noticeable for this particular device.
These initial trials with metallized up to 200MHz Si NW resonators have demonstrated that the Si NWs-in-microtrenches are robust resonators and their anchoring to both the trench walls is indeed reliably self-welded. This verifies that the self-welded anchoring at both the clamping ends does provide sound mechanical rigidity, not only for static loads as shown in the AFM bending experiments [22], but also for the dynamic resonant motions at VHF, and moreover, for operation in their deep nonlinear regimes (which implies longitudinal tension applied to the anchors). With their specifications listed in Table 6-1, the metallized Si NWs resonators have ultrasmall size and mass, excellent
operating frequencies and quite good Q’s, leading to high-performance characteristics as attractive as those of their state-of-the-art top-down counterparts.
Furthermore, for fair comparisons, although the Si NWs diameters are controlled only in a statistical way as compared to top-down lithographically defined features, the main advantage of these as-grown suspended Si NWs is that the fabrication process is easier, faster and much less expensive, as no electron-beam lithography is needed. Since there is no etch process required to suspend the devices, the undercut of the anchoring pads is also avoided, which in principle makes the double clamping more close to semi-infinite and would imply less dissipation through the anchors, as compared to the case where undercut of the anchors is inevitable in the surface nanomachined beam resonators. On the other hand, the top-down method has paramount and relatively more precise control on the number and position of the devices and for multiple devices and arrays, thus realizing rational rather than random device layout.