On-metal UHF-RFID passive tags based on complementary split-ring resonators

The use of complementary split-ring resonators (CSRRs) as radiating elements in low-profile on-metal ultra-highfrequency radio-frequency identification (UHF-RFID) tags is explored in this study. First, the radiation properties of the edgecoupled and the non-bianisotropic (NB-CSRR) versions of the CSRR are studied. The tag design strategy is then discussed in detail. On that basis, a compact (λ0/7 × λ0/7) low-profile (1.27 mm) tag prototype based on the non-bianisotropic version of the CSRR (NB-CSRR) antenna is designed and manufactured to operate in the North-American UHF-RFID band. The experimental results validate the theoretical and simulated behaviour, and exhibit a maximum read range of 6.8 m.


Introduction
Radio-frequency identification (RFID) is a family of technologies which allow identifying objects by using radio waves. To this end, each item is equipped with a device (RFID tag) which stores specific information (generally, an electronic product code, or EPC) and is able to broadcast such information upon an incoming query from a reading device (RFID reader). In the last years, the progressive reduction of the tag cost resulted in a widespread of RFID technology, and the projections indicate further growth in the next years [1]. Among the typical applications of this technology, access control (e.g. electronic toll collection), smart inventory and item tracking are already well established, while other retail-related applications (e.g. smart shelves, automatic payment) are expected for the near future [1].
Passive tags operating in the ultra-high-frequency RFID (UHF-RFID) frequency band (860-960 MHz [2]) have become widely available on the market, since they offer high reading distances, small dimensions and low cost, and do not need batteries to operate. In these tags, an antenna (usually a printed planar antenna) is connected to an integrated circuit (RFID ASIC) which implements the transponder, and contains the specific information regarding the object. When the query signal radiated by the reader is received by the tag antenna, it is delivered to the ASIC, which is able to extract the RF energy to feed its circuitry and generates a series of modulated reflections containing the requested information back to the reader. The reading distance (read range) is in the order of 5-12 m, depending on the tag performance (i.e. antenna efficiency, impedance matching and ASIC sensitivity), the materials of the tagged item and the environmental conditions. Metamaterials and metamaterial-inspired resonators [e.g. splitring resonators (SRRs)] have been recently employed to the design of RFID devices, and showed they can improve the size, the impedance matching and the radiation properties of reader [3][4][5] and tag [6][7][8][9][10] antennas. Also, their complementary counterparts, especially the complementary split-ring resonator (CSRR) and other structures derived from it, were used to obtain compact RFID tags, as follows. In [11,12], modified CSRRs were connected in series to lower the resonant frequency of dipole tag antennas, whereas a CSRR-based inter-layer is used in [13] to shrink the size of a fractal patch antenna for on-metal RFID applications. Nevertheless, the use of complementary resonators as standalone radiating elements in UHF-RFID tags has not been reported. Due to the dipole moments generated at resonance, which satisfy boundary conditions imposed by metallic surfaces, this kind of particle is especially suitable for on-metal tags with very low profile, which still represent a major challenge in RFID technology. It is worth to mention that most of the on-metal UHF-RFID tags found in the literature are based on patch antennas or derived structures [14], and therefore the study of alternative designs is especially interesting.
The paper is organised as follows. First, the radiation properties of CSRRs on metallic surfaces are discussed in Section 2. The RFID tag design strategy is then presented in Section 3, and applied to the design and simulation of a tag prototype. The experimental measurements of the fabricated tag are detailed in Section 4, where a comparison with the state-of-the-art solutions is also provided.

Radiation properties of CSRRs over metallic surfaces
The topologies of the edge-coupled (EC-SRR) and the nonbianisotropic (NB-SRR) SRRs are depicted in Fig. 1. The radiation resistance of the EC-SRR around the first and second selfresonances has been recently calculated analytically as a function of the particle mean radius r 0 [10,15], defined as the arithmetic mean of the inner and outer ring radii. These results also hold true for the NB-SRR, since it presents approximately the same current profile at both the first and second resonances [16]. Furthermore, these results can be directly extended to the case of an ideal (i.e. unbounded and lossless) CSRR, by applying duality [17]. From the well-known Booker relations [18], the radiation conductance around the first and second resonances can be approximated by where Z 0 is the impedance of free space and λ 0 is the free-space wavelength at the operation frequency. Duality suggests that radiation is mainly provided by an axial electric dipole moment and a tangential magnetic dipole moment at the first and second resonances, respectively. In the case of the EC-CSRR, further contribution to radiation arises from the bianisotropy of the structure, which generates a small cross-polar radiation, but its effect on the radiation conductance can be neglected [10,15]. Regarding the tangential magnetic dipole arising at the second resonance, it is oriented along the y-axis in the case of the EC-CSRR [15], as a result of the particle symmetry plane. Conversely, in the NB-CSRR the inner part of the rings is shorter than the outer part, so that the dipole axis is expected to be slightly rotated clockwise, and the rotation angle depends on the geometry of the particle.
Let us now consider the more complex case of a limited and lossy screen, placed at distance h s over a metallic surface. From image theory [19], the structure behaves in the upper half-space (delimited by the ground plane) as a system composed of two parallel complementary screens, with a separation distance 2h s , which presents anti-symmetrical current distribution. Hence, a system of two coupled CSRRs separated by an electric wall is obtained. Coupled resonators theory [20] states that, in this case, the fundamental resonance frequency is lowered (as compared with the uncoupled resonator) when the coupling between resonators is electric, while it is raised when the coupling is magnetic. Since the CSRR generates a strong axial electric dipole moment at the first resonance, coupling is electric, and the first resonance is lowered. Contrarily, the second resonance is lowered by the ground plane because the CSRR exhibits a magnetic behaviour and therefore is magnetically coupled to its image. Additionally, the ground plane reduces the antenna efficiency in both cases, since it introduces several loss mechanisms. First, being the radiation restricted to a half-space, the radiation conductance G rad is halved [18], with respect to an ideal CSRR. Moreover, the currents on the CSRR screen increase (for a given voltage applied across the slot) due to the additional capacitance introduced by the ground plane, which involves an overall increase of the ohmic losses. Finally, the currents induced on the ground plane, or the presence of a dielectric substrate, also contribute to increase losses and reduce efficiency.
The abovementioned behaviour was validated by means of electromagnetic (EM) simulations (CST Microwave Studio), which quantified the input impedance (see Fig. 2), the resonant frequencies and the radiation efficiency (Table 1) of a CSRR etched on a squared copper screen placed at several distances h s on a metallic plane, and driven by a differential port connected at the centre of the external ring. In order to avoid radiation from the patch formed by the finite screen and the underlying ground plane, the four sides of the screen containing the CSRR were shortcircuited to ground by means of vias. The results suggest that, even for very small distances from the metal, the radiation efficiency at the second resonance maintains relatively high values (i.e. η rad = 30% for h s = 1.27 mm).
The simulations did not consider the presence of any dielectric substrate. Obviously, a substrate is necessary in practice, as a support for the top metal layer. The dielectric constant of the substrate can thus be chosen in order to properly reduce the tag size, which is necessary when working at the second resonance because the ground plane causes significant increase in dimensions (roughly 100 mm × 100 mm, for h s = 1.27 mm, which is unpractical for most applications). However, a trade-off arises between final tag size and performance, as already mentioned.

Design strategy
In this section, the design of a low-profile on-metal UHF-RFID tag employing the NB-CSRR antenna at its second resonance is detailed.
The tag layout is depicted in Fig. 3. In this proof of concept, we prioritised size reduction, while maintaining a reasonable read range. Hence, a high-permittivity substrate was chosen, namely the Rogers RO3010 (with relative permittivity ε r = 10.2, loss tangent tanδ = 0.0023 and thickness h s = 1.27 mm). The NB-CSRR was chosen due to its simpler radiation diagram, which is the result of a single radiating dipole moment, and exhibits very low values (ideally zero) of cross-polarisation.
Let us now present the impedance matching strategy between the ASIC and the antenna, which is a critical aspect of passive UHF-RFID tag design. The input impedance of UHF-RFID ASICs is usually modelled by a parallel RC circuit [21], while the antenna can be modelled around resonance by a parallel RLC circuit [22], as shown in Fig. 3c. Due to the very small electrical size of the antenna and to the presence of a nearby ground plane, the radiation quality factor of the system is expected to be high, which leads to high susceptance slopes around resonance [22]. Therefore, a resonance between the antenna and the ASIC is expected just below the CSRR resonance, in the resonator inductive region. At that frequency, the matching level depends on the ratio between G c and G a , which varies as a function of the position φ p of the input port along the slot, as already reported in the case of moving the input port along the SRR ring [10]. Concretely, the NB-CSRR input conductance is raised when moving the port towards the ends of the slot, so that it is possible to obtain good impedance matching without introducing any a matching network. Obviously, the antenna conductance at the centre of the NB-CSRR (which is the lowest value) must be lower than the ASIC conductance to ensure the existence of a position which provides good impedance matching.
To locate the second resonance of the particle to the desired frequency, there are mainly three parameters to take into account.  The first is, obviously, the particle average radius r 0 , which controls the overall lateral dimensions of the particle. Then, the separation d of the slots, which controls the coupling between the resonators, can be used as a design parameter since it offers an additional degree of freedom on the tag dimensions. It is well known that increasing the coupling between the resonators the frequency split of the first and second resonances increases, that is, the first resonance is lowered and the second is raised in frequency. Lastly, the thickness h s of the dielectric also affects the resonance frequencies, since it is the distance between the particle and the metal plane (see Section 2).

Final layout and simulation results
As a proof of concept, a tag working at the North-American UHF-RFID band (902-928 MHz, centred at f 0 = 915 MHz), and based on the Alien Higgs 3 chip transponder (SOT-323 packaged, with power sensitivity P th = −17 dBm and input impedance modelled by G c = 0.66 mS and C c = 0.9 pF), was designed with the aid of EM simulations (CST Microwave Studio), as follows. First, the CSRR dimensions were adjusted to locate the second resonance at 920 MHz, i.e. closely above to the operating frequency, where resonance is expected to occur (see Section 3.1).
The following dimensions were found: L = 45 mm, r 0 = 17.3 mm, c = 1.4 mm, d = 1.8 mm and g = 4.5 mm. Then, the port position was swept until reaching proper impedance matching, which was provided at the position φ p = −20°.
The simulation results, in terms of currents on the screen and electric field inside the slot, are shown in Fig. 4. From the electric field distribution it can be deduced that the equivalent magnetic current in the two rings flows in opposite directions, as expected at the second resonance of the NB-CSRR [16]. However, due to the increase of the resonance frequency caused by the vicinity of the metal plane, the magnetic current presents two maxima, located at roughly ±90°. The simulated input admittance and reflection coefficient, depicted in Fig. 5, shows good impedance matching (−25 dB) at 913 MHz, where the inductive susceptance of the antenna compensates the capacitive behaviour of the ASIC. Finally, the radiation pattern, depicted in Fig. 6, agrees with that of a tangential magnetic dipole moment oriented along the direction φ = 60°, in concordance with theory (see previous section).
The radiation efficiency provided by the simulator is η rad = 12% (ohmic losses on the copper screen account for 62% of the total losses, versus 38% of dielectric losses), resulting in a maximum gain of G r = −4 dB (at θ = 0°). Based on this value, it is possible to predict the maximum tag read range, which is given by [23]    where EIRP is the equivalent isotropic radiated power of the RFID reader, P th is the ASIC read sensitivity and τ = (1 − |s 11 | 2 ) is the power transmission coefficient. By assuming EIRP = 4 W, which is the maximum allowed value in the North-American UHF-RFID band [2], and τ≃1, the predicted read range with Alien Higgs 3 reaches RR = 7.4 m.

Fabrication and measurements
To validate the theoretical and simulated results, the layout described in the previous section was fabricated (Fig. 7a) by means of a printed circuit board (PCB) drilling machine (LPKF-H100). All the measurements were performed with the tag prototype attached to a 20 cm × 30 cm ground plane by means of adhesive tape. The measurements presented below refer to a corrected version of the tag, where the value of r 0 was scaled (1% larger) to correct a frequency shift (15 MHz) detected in the experimental read range. Such a small frequency shift, commonly encountered when fabricating UHF-RFID tags, can be attributed to the fabrication tolerances, and to the simulation accuracies as well. In this case, due to the presence of the metal layer, the tag response also exhibits a strong dependence upon the substrate thickness, which adds further contribution to the frequency shift as compared with a general purpose tag.

Measurement setup
An Agilent N5221A two-port network analyser was used to measure the antenna impedance. To properly excite the antenna, a test fixture based on two semi-rigid coaxial cables soldered together was employed (Fig. 7b). Then, the measured scattering parameters were used to evaluate the differential input impedance, by using the method reported in [24]. The tag read range was measured in a controlled environment, namely a Wavecontrol WaveCell transverse electromagnetic mode (TEM) cell excited by a 50 Ω coaxial cable connected to a vector signal generator (Agilent N5182A), which is able to generate RFID frames at different power levels. The signal backscattered from the TEM cell was sent to an Agilent N9020A signal analyser by means of a circulator, where the tag responses were read. The tag, mounted on a 20 × 30 cm ground plane, was held inside the TEM cell by using a foam support and properly oriented along its main radiation lobe. The tag read range was then evaluated by measuring the minimum interrogation signal power (P min ) necessary to obtain a successful response, and then probing the electric field strength E rms at the tag position by means of an electric field probe (Wavecontrol EFCube). Considering that the average power density S associated to a plane wave is given by and that the power density at a given distance from a transmitting antenna is provided by (where P t is the total transmission power, G t is the antenna gain and r is the distance from the antenna), the read range RR can be obtained by (4) and (5). The result is To obtain the curve of the tag read range, the process was then iterated for each frequency of interest (e.g. at 1 MHz steps).

Experimental results and discussion
The antenna input impedance, depicted in Fig. 5, was measured at the port position φ p = −20° (the RFID ASIC was not mounted at this stage). Reasonable agreement with the simulated values is obtained, though a residual frequency shift is observed (the tag resonates at f 0 = 923 MHz), and the input conductance at resonance (1.6 mS) is higher than the simulated value (0.66 mS). By renormalising the scattering parameters to the impedance of the RFID ASIC, a minimum reflection level of −8 dB (consequently, τ = 0.84) is obtained (Fig. 5), which involves a 8% reduction of the peak read range, as compared with conjugate matching. The experimental read range is depicted in Fig. 8. The tag response is centred at the frequency f 0 = 923 MHz, and the maximum read range is 6.8 m, which is in very good agreement with simulations when the mismatch correction mentioned above is applied. In light of its novelty and performance, the proposed approach is an interesting alternative to the commonly employed patchderived structures. In fact, as can be seen in Table 2, the presented tag is competitive against the state of the art of very-low-profile on-metal UHF-RFID tags. However, further research on the subject is needed to avoid using high-permittivity substrates, and/or to further reduce the tag thickness or lateral dimensions.

Conclusion
The radiation properties of CSRRs placed at very small distances from a metallic plane have been studied in this work. Based on duality, qualitative predictions of the particle behaviour were provided and then validated by EM simulations, which suggested that the CSRR presents acceptable radiation efficiency near its second resonant frequency even for very low distances from the ground plane. Based on these results, the design strategy of a passive UHF-RFID tag based on such antenna was then presented, where it was shown that good impedance matching between the RFID ASIC and the NB-CSRR antenna could be obtained without the need of any matching network. As a proof of concept, a compact (≃λ 0 /7 × λ 0 /7) and low-profile (1.27 mm) tag based on the NB-CSRR antenna was designed and fabricated to operate on metal in the North-American UHF-RFID band. The experimental read range is in good agreement with simulations, and reaches 6.8 m, thus validating the suitability and competitiveness of the proposed approach for building on-metal tags with reduced thickness.

Acknowledgments
This work was supported by MINECO-Spain (project TEC2013-40600-R COM-SEN-RFID), FEDER funds and Generalitat de Catalunya (project 2014SGR-157). Ferran Martín is in debt to ICREA for supporting his work and by awarding him with an ICREA Academia Award.