Printable electronics (PE) is an advanced manufacturing technology that is of significant interest to a large range of industries, from consumer goods, electronics, aerospace, automotive, pharmaceutical, biomedical, to textiles and fashion [1
]. It offers an attractive alternative to conventional circuit manufacturing by enabling lower-cost, maskless, and rapid production of customized electronic devices [6
]. PE is compatible with a wide range of substrates, as long as they are not porous and can resist all fabrication steps, including pre- and post-printing processes [7
]. In addition, various kinds of conductive, semi-conductive, and dielectric inks are now commercially available. Therefore, PE allows the realization of unique electronic components that can be bent, twisted and stretched, all while retaining their electrical properties [8
]. In recent years, the development of various contact- and non-contact printing technologies, such as flexography, gravure, screen- or inkjet-printing, has advanced significantly [6
]. Post-printing processes also play a key role in the manufacturing of PE devices. The most commonly used sintering approaches are conventional thermal annealing, electrical sintering, microwave, and photonic sintering by either continuous-wave laser irradiation or high-power flashing lamps [7
]. While the spatial resolution and definition of the device are related to the printing method, the quality of the electrical properties of the printed devices is directly related to the post-printing process [12
]. Particularly, the solid and uniform dielectric or metallic tracks from the printed pattern are obtained during this step.
The quality of PE devices can be evaluated using different types of microscopy, such as atomic force microscopy, scanning electron microscopy or optical microscopy [13
], which are well-established tools for analyzing the surface morphology of materials. Nevertheless, these techniques are expensive, slow, and allow limited surface area observation. Other types of characterization techniques, such as crystallography analysis, thermography, elecro- or photo-luminescence, are also time-consuming and require special conditions, such as vacuum or helium environments, to avoid noise and damage [15
]. The electrical conductivity of printed traces in flexible PE circuits is assessed using conventional methods drawn from the electronics industry, e.g., the flying probes or four-point probe method (4PP). However, these techniques cannot be envisioned for high-volume roll-to-roll (R2R) printing since in-line contact methods are not compatible with continuous manufacturing tools. Thus, a non-contact conductivity characterization method is necessary.
Traditional graphic art printing or off-set printing used in the manufacture of full-color magazines, posters, packaging, etc., evaluates print quality using a color control bar (GATF Standard Offset Color Bar) [19
]. Using a densitometer or a spectrophotometer, these bars allow for accurate determination of ink density, dot gain, and screen angle accuracy. Generally, the color control bars are printed away from the immediate image area, and are often cut off or hidden during final assembly. Similarly, for PE production, an in situ quality control characterization technique has to be developed. Time-domain spectroscopy (TDS), using electromagnetic terahertz (THz) radiation, i.e., for frequencies ranging from 100 GHz to 10 THz, is a powerful tool that allows non-destructive characterization, and which is very sensitive to the conductivity of matter [20
]. THz waves have previously been used to characterize carbon printed ink with the THz imaging method [22
]. However, for high volume production, such approach is time consuming and may require complicated data analysis to efficiently recover the conductive property of the printed devices. Alternatively, THz engineered structures, such as metamaterials [23
], can exhibit a strong response in transmission- or reflection-type geometries with a high dependency on material conductivity [24
]. Therefore, it can provide a straightforward sensing tool to retrieve the conductive property of the printed ink. Already, THz metamaterials printed by inkjet [26
], digital aerosol jet [30
], laser printing [31
] or electro-hydrodynamic jet [32
] printing have been reported, allowing for rapid fabrication of THz metamaterial-based sensors and functional THz devices using PE methods [27
In this work, a THz engineered resonance structure has been developed as a quality control bar to probe the post-printing manufacturing process of PE devices. Our objectives were to determine the transmission resonant behavior of a control bar using THz waves as a function of ink conductivity and to link the THz frequency conductivity with the static conductivity of printed devices that are manufactured simultaneously (i.e., with the same sintering condition). As illustrated in Figure 1
, we have performed a comparative study between THz inspection of a resonant printed structure against conventional conductivity measurement methods on a printed structure, i.e., using multimeter (MM), four-point probe (4PP) and atomic force microscopy (AFM). Our THz measurements are well-correlated with the non-resonant printed structure conductivities and confirm the ability to determine the quality of the post-printing manufacturing process of PE devices by THz inspection of a simple control bar showing a distinctive response in the THz frequency range. To retrieve the resonance response of our control bar, standard terahertz time-domain spectroscopy (THz-TDS) was utilized. In addition, the well-known THz transmission method was compared through a novel dual-wavelength THz spectroscopy (DWTS) analysis. We show that DWTS determines the conductivity of the PE device using a single scan measurement. Additionally, our method does not rely on THz phase-sensitive measurements, and is therefore ideally suited for next-generation low-cost THz emitters and sensors [35
] and opens the door to contactless in situ quality control of PE devices.
3. Results and Discussion
Five VPP samples with different conductivities were characterized by the THz-TDS described above. The conductivity of each sample was controlled by varying the sintering time. One of the samples (non-sintered) was not sintered by the lamp, but was slightly sintered during the printing step, since the chuck was held at a constant temperature of 60
C. Figure 3
a illustrates the normalized transmission amplitude of the different VPP samples, which were obtained from Equation (1
). A dip in the transmission is observed due to the generation of a vortex beam at 0.22 THz, as expected [37
]. As mentioned previously, a higher resonance response (i.e., which translates to a lower transmission at 0.22 THz) indicates a sample with higher electrical conductivity.
To validate the accuracy of THz sensing of vortex plates as a function of material conductivity, we performed finite difference time domain (FDTD) simulations using the Lumerical software. Linearly polarized waves and perfectly matched layer boundary conditions were used in the simulation.
b shows the simulated transmission spectra of VPPs with defined and uniform conductivities of a hypothetical printed metal. We placed VPP in the air in order to avoid Fabry–Perot resonances from the substrate. We can observe three transmission dips; the strongest one at 0.265 THz represents the central frequency of VPP. Compared to experiments, the red shift of the central frequency is explained by the absence of the PET substrate.
The simulation and experiment differ in the degree of transmission difference as a function of metal conductivity. We attribute this difference to the perfect reading of the central vortex information in the simulated case. Essentially, the photoconductive antenna reads a spatially integrated range of information containing the central intensity part of a donut shaped beam, together with a large contribution from its wings. Nevertheless, the numerical simulations are in good agreement with experimental findings.
c gives the measured conductivity of five samples using three different methods: 2MP, 4PP, THz-TDS and DWTS as a function of sintering speed. The 4PP method was performed on the patch samples, while 2MP, THz-TDS and DWTS measurements provide the corresponding conductivity results from the VPP samples. The function of the value of the dip in transmission against the conductivity of VPP was also simulated, as shown in Figure 3
d. It is important to note that this function clearly reveals the extremely high sensitivity of THz wave sensing for low conductivity samples (e.g., below
S/m, the blue dotted region in inset). Above this conductivity value, the dip in transmission exhibits less sensitivity, with an almost saturated behavior (i.e., closer to a perfect metal resonance).
To compare the performance of THz-TDS and 4PP, the THz transmission amplitudes at 0.22 THz were calibrated to the expected conductivity values obtained from 4PP. Since the 4PP measurements cover a limited range of conductivity, from
S/m, a simple calibration using a linear fit was chosen (in agreement with inset of Figure 3
d, with the non-sintered sample as the starting point. In Figure 3
c, the similar increases in conductivity behavior as a function of sintering exposure time for the measurements taken by THz-TDS and 4PP is observed. More importantly, all sintering conditions are well discriminated by THz measurements, whereas 4PP failed in differentiating the three lowest conductivity conditions (i.e., <
S/m), as well as the two highest conductivity conditions (i.e., >
S/m). In addition, we repeated the measurements ten times for each method and calculated the standard deviation. Interestingly, THz-TDS exhibits better repeatability than the conventional 4PP method. We attribute this difference to the contactless nature of the THz method: 4PP can locally damage the ink surface and may render repeated measurement less accurate.
In the second step, using the data obtained from THz-TDS measurements, we analyzed the sample signal by the DWTS method. The two frequency ranges were 0.195–0.244 THz and 0.615–0.664 THz, for the signal and reference, respectively (see Figure 2
b). In order to perform the measurement in ambient conditions, the reference frequency range was chosen to avoid the water absorption lines that can occur due to ambient humidity. Similarly to THz-TDS transmission data, the integral values from DWTS were normalized and calibrated with respect to the retrieved conductivity using the 4PP method. The behavior follows the expected static conductivity, but more importantly, the repeatability is four times better than the conventional 4PP method.
In the final step, we review the analysis done on the patch versus VPP samples using the various methods described previously. Table 1
summarizes the obtained results. In order to establish a comparative measurement performance, we carried out several resistance measurements at different locations for the patch and V-shaped antenna and present their relative standard deviation (RSD). As mentioned previously, the 4PP and multimeter retrieved the resistance on the patch. To clearly validate that VPP conductivity is linked to the patch conductivities, 2MP were also used to evaluate the VPP resistance directly. It has to be mentioned that, due to the extremely small effective volume of VPP unit cell, the 2MP method can easily over- or underestimate the conductivity (e.g., conductivity dependency on sample volume, as shown in Equation (3
)). However, the 2MP measurements confirmed the good agreement between the sintering exposure time for the patch and VPP samples together.
In order to confirm the provided conductivity measurements, the evolution of the sintering of Ag ink was studied using AFM analysis at five different sintering stages. The last row of Table 1
depicts the printed ink surfaces after sintering. The non-NIR-sintered sample (NS) showed poor contact between Ag NPs, resulting in the lowest conductivity (
S/m). The sample with the shortest annealing time (0.03 s/mm) depicted the next stage of the sintering, necks began to grow between NPs prompted by surface energy minimization. With a longer annealing time of 0.05 s/mm, the NPs get more compact and the printed structure densifies. The slight increase of annealing time to 0.07 s/mm led to a further increase in conductivity. The longest annealing time (0.2 s/mm) led to the highest density and the highest conductivity (
S/m). According to AFM observations of the surface morphology of the samples, the obtained samples were consistent with the sintering parameters and measurements of the conductivity with different techniques.
As can also be seen in Table 1
, as expected, the measurements provided by a conventional multimeter were the least precise since the probes of the multimeter easily break the surface of the patch after contact. Meanwhile, the micro-probe provides a safer way to avoid destroying the sample surface. The average conductivities measured with the different techniques are in the same range, and have similar behavior as a function of the sintering time. It should be emphasized that the trend in electrical static conductivity measurements on the printed patch and the VPP using the different techniques are all in good agreement. This confirms the feasibility of characterizing the variability in ink conductivity during mass production of PE devices simply by reading a test structure. Finally, the best RSD for repeatability was obtained for DWTS and THz-TDS.