Babinet’s principle for quality control of printed electronics

Towards in-situ quality control of conductive printable electronics: a review of possible pathways

“…Visual inspection is commonly used for PE quality control, and more specifically, with optical microscopy (OM), which is a widely used method for evaluating traces or printed devices. Standard compound OM is a common technique as it is relatively simple to capture a surface [23,31,32,36,40,45,82,[87][88][89], while confocal microscopy is used to provide threedimensional (3D) information at the micrometer level [40,59,67,82,83,87,90].…”

“…Microscopic images are usually taken at each stage of production, from ink development [31,32,40,87], substrate preparation [82], the deposition process [23,59,67,83,88,89], to sintering/drying [32,33,36,45]. This is particularly useful for characterizing the nature of crack formation [36] or for detecting different types of defects such as impacted droplet boundaries, the wetting phenomenon, irregularities, pinholes, etc [40,90].…”

“…Similarly, Zhuldybina et al [83] studied the printed metal checkerboard (MCB) pattern to extract lattice parameters using OM images with image postprocessing, as shown in figure 14. The designed and printed images are shown in figures 14(a) and (b), respectively.…”

“…To control the quality of conductive PE layers, Zhuldybina et al [83,114] iintroduced a simple strategy using a THz resonant structure as a quality control bar (QCB) for printing and characterized it by THz time-domain spectroscopy (THz-TDS). The main part of the THz-TDS system is a femtosecond (fs) laser, which, through an emitter, transforms the fs pulse into a picosecond THz pulse [115].…”

“…Characteristics AODE [96] LED [102] Non-linear OM [71] THz imaging [117] ECT [93,104] ST [106,107] THz-TDS [83,114] SAXS [99] Thickness extension of this capability is the implementation of artificial intelligence (AI) assistance, in particular, machine learning algorithms (MLAs) that can selflearn a production process by comparing the results obtained with a defined threshold for electrical performance and/or defect size. This type of machine learning has already been used in a similar way for pass and fail cases [122].…”