SAW RFID for wireless temperature measurement
The peak width of an SAW RFID device in the time domain is inversely proportional to its bandwidth in the frequency domain, as the time-bandwidth product of the device is fixed. A narrow peak width facilitates a large coding capacity, while a wide bandwidth enhances measurement accuracy. The electromechanical coupling factor k2 of the SAW substrate determines the upper limit of the achievable bandwidth. Consequently, selecting a SAW mode with a high k2 is advantageous for designing high-performance SAW RFID devices. Shear-horizontal (SH) modes on single-crystal substrates such as 32°YX-LiTaO3, 41°YX-LiNbO3, and 64°YX-LiNbO3 exhibit large k2 values but a leaky nature on free surfaces. SAW RFID devices for wireless sensing applications typically require long acoustic propagation paths (often exceeding 1 μs) to generate sufficient time delay for coordination with the reader. A significant portion of the SAW energy dissipates during propagation on the free surface. Although the k2 of Rayleigh modes on YZ-LiNbO3 and 128°YX-LiNbO3 is considerably lower than that of SH modes, they are more suitable for wireless applications due to their non-leaky nature. Additionally, Rayleigh modes on YZ-LiNbO3 and 128°YX-LiNbO3 have a large and stable temperature coefficient of time delay. Thus, the time delays and phases of SAW RFID devices on these substrates have high and linear sensitivities to temperature, which is an outstanding characteristic for accurate temperature measurement. The SAW RFID device on 128°YX-LiNbO3 has a larger critical dimension than one on YZ-LiNbO3 operating at the same frequency, because the phase velocity of the Rayleigh wave on 128°YX-LiNbO3 (approximately 4000 m/s) is faster than that on YZ-LiNbO3 (approximately 3500 m/s). The 128°YX-LiNbO3 substrate was selected in this study to reduce the fabrication difficulty.
Aluminum is widely employed in SAW RFID devices for its high electrical conductivity and low density. These properties enable aluminum electrodes to achieve relatively low ohmic loss and to excite high-frequency SAWs. The normalized thickness (ratio of electrode thickness to wavelength, h/λ) of aluminum electrodes in SAW RFID devices typically ranges from 0.63 to 3.7%13,15,17,31,32. A normalized aluminum thickness of 1.67% was selected in this study, which falls within the common range. The aluminum electrode used in this work was doped with 2 wt% copper to enhance its conductivity.
RMSC reflector with high reflectivity
It is essential to avoid mechanical reflection during the design of the RMSC reflector, as it causes undesired reflected peaks in the time domain. This reflection is primarily influenced by the pitch of the RMSC structure. To identify a suitable pitch, the reflection characteristics of RMSC reflectors with varying pitches were simulated using the model illustrated in Fig. 1S. The results indicate that when the RMSC pitch exceeds half the wavelength, strong impedance mismatch leads to high-amplitude reflection peaks at both edges of the RMSC, which can degrade the amplitude of the main peak and coding capacity of the device. When the RMSC pitch equals half the wavelength (satisfying the Bragg condition), the acoustic wave is trapped within the RMSC strips owning to the strong internal reflection. The trapped energy gradually leaks out, leading to a roll-off in the time domain. Conversely, for pitches smaller than half the wavelength, the impedance mismatch is weak, resulting in weak reflection. Moreover, a shorter pitch of the RMSC strip reduces the total width of the RMSC, causing the left and right reflectors to merge with the main peak. Narrower electrodes in RMSC designs produce weaker reflections caused by acoustic impedance discontinuities, thereby enhancing device performance. However, reducing the electrode width increases ohmic loss and introduces fabrication challenges due to the smaller critical dimension. After balancing ohmic loss, manufacturing constraints, and reflection suppression requirements, an RMSC electrode pitch of 3/8 wavelength was selected as the optimal compromise.
The performances for RMSCs with an aperture of 70 wavelengths were simulated as a function of strip number using the structural parameters shown in Table 1. The RMSC pitch is set to 3/8 wavelength to avoid Bragg reflection, and the gap between port 2 and port 3 is set to half of its pitch to minimize resistance and possible bulk wave conversion, while maintaining a consistent critical dimension, as shown in Fig. 2a. The parasitic effects imposed by strip resistance and capacitance between connection strips were included in the simulation. The simulation results of average transmittance, reflectance, and loss for Type1 RMSCs across 413–453 MHz were calculated, as shown in Fig. 2b. For ideal RMSCs, where neither capacitance effects nor resistance effects are included in the numerical simulations, the energy loss is minimal and independent of the strip number. The strip number only determines the ratio of reflected to transmitted energy. The maximum reflectance and minimum transmittance occur simultaneously when the strip number is 40. The RMSC with this optimal strip number achieves near-total reflection over a wide frequency range, with only 0.04 dB of SAW energy being lost or transmitted.
a Schematic of Type1 and Type2 RMSCs. b Simulated average transmittance, reflectance, and loss for Type1 RMSCs as a function of strip number. c Simulated average reflectance for RMSCs with different connecting strip types as a function of strip number. d Simulated average reflectance for RMSCs with different apertures as a function of strip number. e Measured transmittance, reflectance, and loss of the RMSC with optimal strip number as a function of frequency compared to the ideal model and the model considering the parasitic elements
The ideal RMSC can achieve near-total reflection. However, the ohmic loss and stored charges damage its performance. Ohmic loss occurs when SAWs propagate through the RMSC reflector, as currents induced in the metal electrodes are dissipated due to the inherent resistance. This phenomenon was simulated by incorporating parasitic resistances into the finite element model in this manuscript. After being excited by the incident wave, the induced charges distribute across the entire metal strips, including the connecting segments. Charges on the connecting strips generate SAWs with wavevectors perpendicular to these strips. However, the wavenumber of these SAWs differs from that of the incident wave because the phase velocity of acoustic waves varies with the propagation direction. The resulting wavelength mismatch with the periodic metal strip pitch prevents these weak SAWs from constructively interfering and amplifying. Only a small fraction of the charges on the connecting strips regenerate SAWs, while most are stored due to capacitances between adjacent connecting strips. This phenomenon is simulated by incorporating parasitic capacitances into the finite element model in this manuscript.
The parasitic resistances of strips and parasitic capacitances between connecting strips have a significant impact on the performance of RMSCs. The ohmic loss power for an acoustic wave is proportional to the acoustic wave power and total resistance. When the RMSC satisfies the interference condition, the entire incident wave is converted into a reflected wave, with no excited transmitted wave. If the RMSC reflector has slightly fewer strips than required for perfect constructive interference, the amplitude of the reflected wave is weaker. Consequently, the ohmic loss power of the reflected wave decreases due to the reduced induced current and lower associated parasitic resistance. For the RMSC suffer from large ohmic loss, this reduction in ohmic loss may outweigh the decline in the excitation efficiency of the reflected wave caused by the deviation from optimal interference. This trade-off can potentially reduce the number of strips needed to achieve maximum reflectance. In contrast, parasitic resistances do not significantly influence the strip number required for minimum transmittance. This is because any transmitted wave resulting from a deviation from perfect constructive interference cannot be fully attenuated by ohmic loss. The storage charges reduce the electro-acoustic conversion efficiency of the RMSC, which is equivalent to a decrease in the electromechanical coupling coefficient k2 of the SAW substrate near the RMSC region. The reduction in k2 caused by the parasitic capacitances narrows the phase velocity difference between SW and the AW. As a result, additional strips are needed to provide sufficient propagation length for achieving a 90° phase difference between the modes. Therefore, parasitic capacitances increase the strip number required to meet the interference condition. Affected by both factors, the reflectance of the practical RMSC is less sensitive to variations in the strip number compared to the ideal RMSC. Moreover, the maximum reflectance and minimum transmittance no longer always occur together. The lowest transmittance is observed at a strip number of 50, whereas the highest reflectance occurs at a strip number of 40, with the combined lost and transmitted energy reaching 1.84 dB.
To reduce parasitic capacitances and SAW generation due to the connecting strips, the connection strips were alternately placed, one at the top and one at the bottom (Type2 RMSC in Fig. 2a). The distances between connection strips in Type2 RMSC are greater than those in Type1 RMSC, resulting in lower connection strip capacitances. The maximum simulated reflectance increases from 65.4% for Type1 RMSC to 69.3% for Type2 RMSC, as shown in Fig. 2c. The simulated average reflectance for type2 RMSCs with apertures of 40 wavelengths, 70 wavelengths, and 100 wavelengths was calculated, as shown in Fig. 2d. The RMSC with a short aperture has smaller RMSC strip capacitance and is susceptible to parasitic capacitances from connection strips, requiring more strips to achieve the highest reflectance. Additionally, the RMSC with a shorter aperture has smaller ohmic loss and higher maximum reflectance. The RMSC with an aperture of 40 wavelengths achieves a maximum reflectance of 83.3% at 45 strips. Further shortening the aperture improves reflectance but may introduce significant diffraction loss in SAW RFID.
A Type2 RMSC with 45 strips and an aperture of 40 wavelengths was fabricated for experimental validation, with its transmittance, reflectance, and loss spectra shown in Fig. 2e. The near-zero measured transmittance aligns closely with simulations of the parasitic-inclusive model. However, a 0.33 dB discrepancy in energy loss was experimentally observed, potentially attributable to SAW radiation from connection strips, fabrication-induced and environment-induced errors, and bulk wave conversion. As discussed previously, the charges distributed on connecting strips excite weak SAWs. This phenomenon is not included in the simulation and could contribute to the discrepancy. Additionally, the electrode thickness deposited via electron-beam evaporation was non-uniform across the wafer, leading to slight variations in ohmic loss among devices. Exposure to airborne dust and moisture during testing changed the propagation loss of surface acoustic waves on the free surface. Therefore, fabrication-induced and environment-induced errors are potential factors for the observed discrepancy. Moreover, bulk wave conversion occurring at the left and right edges of the RMSC strips cannot be fully captured in the simulation, which could also contribute to the discrepancy. The simulation results of the ideal model deviate from experimental data, rendering this model unsuitable for RMSC design optimization. Crucially, the RMSC maintains approximately 77.8% average reflectance across 393 to 473 MHz, resulting in only 1 dB of energy loss. The high reflectance effectively reduces the energy loss of the device and enhances the interrogation distance of the wireless SAW RFID system. Additionally, the wide frequency range contributes to reducing the peak width of the time domain response, thereby expanding coding capacity and increasing measurement accuracy. These characteristics make the RMSC suitable for application as an SAW RFID reflector.
To compare the performance of RMSC with traditional reflectors, time domain responses of open-circuited (OC) and SC reflectors with varying electrode numbers were simulated using the models shown in Fig. 2S. The results demonstrated that when the electrode number exceeded 20 for the OC reflector or 40 for the SC reflector, strong internal reflection between electrodes occurred. Further increase in the electrode count did not significantly enhance the peak amplitude. The reflectances of the SC reflector with 20 electrodes and the OC reflector with 40 electrodes were both below 21%, a value significantly lower than the reflectance achieved by the RMSC reflector.
SAW RFID using RMSCs with maximum reflectivity
A novel SAW RFID architecture employing RMSC technology (referred to as RMSC SAW RFID) is proposed in this section. Two optimized RMSCs, serving as high-performance reflectors, are positioned on both sides of a single-electrode-type bidirectional IDT, as illustrated in Fig. 3a, b. The IDT with 8 electrode pairs operating at the fundamental harmonic frequency can achieve an available bandwidth of more than 80 MHz. The designed SAW RFID exhibits three distinct high-amplitude peaks in the time domain. SAWs excited by the IDT propagate toward RMSCs and are then reflected back to the IDT. A portion of the reflected signals is converted into electrical signals, forming the first and second peaks with propagation distances of 2L1 and 2L2, respectively. The remaining waves propagate and are reflected by another reflector, generating the third peak with a propagation distance of 2L1 + 2L2.
a Schematic layout of the SAW RFID with RMSC reflectors. b Photo of the fabricated chip. c Schematic of impedance matching circuit for the SAW RFID. d Measured Smith chart of the unmatched and matched SAW RFIDs. e Measured frequency domain response of the unmatched and matched SAW RFIDs. f Measured time domain response of the unmatched and matched SAW RFIDs
The impedance mismatch between the IDT and the 50 Ω system is another critical factor contributing to energy loss. Increasing the number of IDT finger pairs helps align the IDT impedance closer to 50 Ω. However, a larger number of finger pairs significantly reduces the device’s bandwidth. Additionally, this increase leads to higher internal reflections, potentially altering the peak shape in the time domain. To address this issue, an impedance matching circuit is implemented to eliminate the impedance mismatch at the IDT interface, as shown in Fig. 3c. The matched IDT exhibits an impedance closer to 50 Ω, with a wide operational bandwidth, as shown in Fig. 3d, e. This circuit reduces the input reflection at the IDT to 50 Ω system interface by approximately 20 dB, evidenced by the reduced peak amplitude at time t = 0 in matched versus unmatched devices in Fig. 3f. The bandwidth of SAW RFID is over 80 MHz. However, the relative bandwidth of commonly used compact antennas is typically less than 10%. In practical wireless measurements, the available bandwidth of SAW RFID systems is often constrained by antenna limitations. To assess performance under real-world conditions, the operational bandwidth was set to 40 MHz in this study. The peak amplitudes in the time domain corresponding to a 40 MHz operational bandwidth are listed in Table 2. The peak amplitude in the time domain employed in this study aligns conceptually with the negative value of insertion loss referenced in prior work7. Insertion loss is a frequency-domain parameter that quantifies signal energy attenuation. However, its effects are directly observable in time domain waveforms due to the Fourier relationship between domains. In SAW RFID systems, insertion loss is often used to characterize the peak amplitude in the time domain. In this manuscript, we utilize the peak amplitude in the time domain instead of insertion loss to reduce ambiguity. Impedance matching enhances the amplitude of the first and second peaks by approximately 7.4 dB. Concurrently, matching reduces the unconverted components in RMSC-reflected SAW echoes, resulting in diminished improvement for the third peak amplitude (3.4 dB). There is a spurious signal between the 2nd and 3rd peaks due to a wave running twice the distance L1.
A comparison between our work and state-of-the-art SAW RFID devices with high peak amplitudes is provided in Table 3. Short-circuited (SC) gratings satisfying Bragg conditions are widely applied in conventional SAW RFID designs. The reported maximum peak amplitude for the SAW RFID with SC gating reflectors and a bidirectional IDT is around -30 dB. The peak amplitude of this type of SAW RFID can be further improved by replacing the bidirectional IDT with a single-phase unidirectional transducer. However, this reduces the relative bandwidth to 2.13%, which compromises the coding capacity and measurement accuracy. Another SAW RFID structure, termed the connected IDT (CIDT) SAW RFID, achieves low energy loss as reflectors are unnecessary in this structure. The reported maximum peak amplitude for the CIDT SAW RFID is around −18 dB. The RMSC-based SAW RFID developed in this work achieves a peak amplitude of −10.6 dB, which is at least 7.4 dB higher than that of conventional SAW RFID devices. Consequently, wireless SAW RFID systems utilizing the proposed RMSC structure can achieve a reading range at least 1.53 times greater than the same systems employing other types of SAW RFID, as a 12 dB reduction in energy loss doubles the readout distance.
The temperature characterization of the fabricated RMSC SAW RFID was measured. Time delay combinations and phase combinations extracted from the three reflected peaks are plotted as a function of temperature from −20 to 90 °C in Fig 4a. Here, τi and φi (i = 1,2, or 3) are the time delays and phases of the three reflectors. τ31 denotes τ3−τ1, φ31 represents φ3−φ1, and φ3221 is defined as (φ3−φ2)−(φ2−φ1). Substrate thermal expansion and temperature dependence of crystal stiffness constants induce a linear increase in propagation length and decrease in the phase velocity with rising temperature, consequently increasing time delays linearly. Both time delays and unwrapped phases exhibit linear thermal dependencies (ordinary determination coefficients R2 > 0.9999). The scattering loss of SAW arises from thermal lattice vibrations that intensify with temperature, thereby reducing the reflected peak amplitude. Longer propagation paths exhibit higher susceptibility to this effect. As temperature increases from −20 to 90 °C, the first, second, and third peak amplitudes decrease by 1.3, 1.7, and 2.9 dB, respectively, as shown in Fig 4b.
a Measured τ31, φ3221, and φ31 as a function of temperature from −20 to 90 °C. b Measured amplitudes of first, second, and third peaks in the time domain. c Fluctuation of the temperature signal calculated from the unaveraged measured signals τ31, φ3221, and φ31 of the SAW RFID, and PT100 under 0, 30, and 60 °C temperature load
Figure 4c compares temperatures determined by RMSC SAW RFID against PT100 reference values. Testing at 0 °C, 30 °C, and 60 °C yielded 500 datasets per temperature under low signal-to-noise ratio conditions. Despite exceeding the unambiguity measuring range of φ3221 and φ31, the 2π phase ambiguities were effectively eliminated by a multistep evaluation scheme33, as evidenced by error-free measurements. The 6σ uncertainties of the temperature measurements are summarized in Table 4. Experimental results indicate that phase estimations exhibit significantly higher precision than time delay estimations. Chamber temperature fluctuations dominate the internal temperature uncertainty of φ31, making φ31’s intrinsic error negligible. Consequently, the temperature uncertainty of φ31 matches the PT00 reference. Omission of higher-order terms in the estimation model introduced minor bias (defined as the average deviation between estimated and PT100 reference temperature).
The observed characteristics of high linearity, low loss across a wide temperature range, and closed experimental-simulation matching demonstrate the excellent suitability of RMSC SAW RFID technology for temperature measurement.