1. INTRODUCTION

Metasurfaces have garnered a great deal of attention because of their unique ability to tailor the properties of transmitted and reflected radiation. Recent work in the area of metasurfaces has shown that such structures can be used to enable a wide range of complex beam shaping operations, providing control over the amplitude, phase, and polarization of the incident radiation [1–24]. One of the more intriguing applications is in the area of image formation, often accomplished using holography [17–21], where the phase of the reflected optical beam can be altered by making changes to individual unit cells, enabling the formation of the desired image. Much of the work in metasurface holography has relied on the use of resonant structures fabricated using metals that are good plasmonic materials in order to minimize loss at the resonant frequency. Thus, gold and silver are often the metals of choice. In the terahertz (THz) spectral range, typically defined as extending from 0.1 to 10 THz [25], the choice of plasmonic materials is much broader and includes not only all conventional metals, but also a wide range of unconventional metals. These include graphene [26,27], liquid metals [28], and heavily doped semiconductors [29,30] that can have DC conductivities that are one to two orders of magnitude lower than that of gold. By varying the conductivity of each individual resonant structure within an array, we can vary the associated loss, which corresponds to a change in the transmissivity at the resonant frequency. Creating such a spatial variation can be accomplished using a variety of techniques. For example, a nine gray-level array can be fabricated using conventional microfabrication techniques, although it would require repeating the same basic process nine times. Alternatively, we have shown that inkjet printing allows for the fabrication of a spatially varying conductivity that can be performed within a single fast print cycle using a combination of silver and carbon ink, although there are limits on the spatial resolution [31]. In earlier work where the conducting medium was continuous, a rapid spatial variation in conductivity would have been averaged out by the propagating surface plasmons [31]. By creating electrically isolated unit cells, large variations in conductivity can be made to adjacent unit cells with reproducible variation in the resonance properties observed in transmission.

2. MATERIALS AND METHODS

THz transmission spectra were obtained using standard time-domain spectroscopy [32,33]. We used $⟨110⟩$ ZnTe optical crystals for both THz generation and detection, stimulated by an ultrafast Ti:sapphire laser [35]. The broadband, linearly polarized THz radiation generated by the emitter was collected and collimated by an off-axis parabolic, such that the radiation was normally incident on the samples. A second off-axis parabolic mirror placed after the sample was used to focus the transmitted THz radiation onto the detection crystal for measurement via electro-optic sampling. For THz imaging [36], we used three separate narrowband polarized electronic THz sources operating at 0.1, 0.15, and 0.3 THz. This radiation was collimated and imaged onto a THz focal plane array using low-loss THz lenses.

The metasurface pattern was printed using a commercial inkjet printer using a combination of conductive and resistive inks [31] to yield unit cells that had differing conductivities. The metasurface samples were printed using a color inkjet printer, which allowed for a minimum print resolution of $\sim 25\mathrm{\mu m}$. The silver ink sintered at room temperature immediately after printing, regardless of the carbon ink content, yielding a final pattern that required no additional post-processing steps. A single solid film using only silver ink yielded structures with a DC conductivity of $5\times {10}^7\mathrm{S}/\mathrm{m}$, which was nearly the same conductivity as that of bulk silver [31]. The modulation levels in terahertz transmission were determined using test structures composed of arrays of identical resonant dipoles. These test structures were characterized using conventional THz-TDS measurements taken along the length of dipoles. In case of 2- or 3-color metasurfaces, the modulation levels were determined for each dipole component.

3. RESULTS AND DISCUSSION

In this investigation, we measure the THz transmission properties of arrays of unit cells that contain a single dipole, a cross-shaped resonator, or a combination of three unit dipoles in which equivalent meta atoms in adjacent cells can have different conductivities. The resonant structures are designed to have a lowest-order resonance at 0.1, 0.15, or 0.3 THz and are interrogated using narrowband polarized radiation at those frequencies. We used a combination of THz time-domain spectroscopy [32–34] and THz imaging [35,36] to investigate these structures. For the latter technique, we used multiple narrowband polarized THz sources in conjunction with a THz focal plane array.

In Fig. 1(a), we show a photograph of a metasurface composed of planar dipoles printed on thin, low-loss polyethylene terephthalate (PET) sheets, where all of the dipoles are oriented along the same direction. Each dipole in this image is printed using a combination of silver and carbon inks, with the silver fraction varying between 50% and 90% from left to right. Over this range, there is no observable difference between the dipoles visually. However, the difference in conductivity leads to noticeable differences in the THz transmission. In Fig. 1(b), we show the THz transmission spectra for uniform arrays of dipoles that were printed with different fractional silver contents. The resonant frequencies associated with the dipole are related only to its length along the dipole axis and are independent of the silver–carbon ratio. Furthermore, the depth of the lowest-order resonance is linearly related to the fractional silver content, but only over the range mentioned above (see Supplement 1 for more details). For the imaging results discussed below, only the lowest-order resonance was considered. In Fig. 1(c), we show a schematic diagram of the THz imaging scheme. In contrast to previous metasurface imaging demonstrations, images are formed based on the amplitude of the transmitted THz radiation. As would be expected from the resonance properties of rods (dipoles) that are effectively one-dimensional, we only observe well-defined images when the incident THz radiation is polarized parallel to the dipole axis.

 

Fig. 1. Design, transmission properties and imaging scheme for THz metasurface. (a) (left) Optical image of an inkjet printed dipole array with a red bar (500 μm length) to show dimensions. The horizontal and vertical gaps between dipoles was 300 μm; (right) optical image of a $24\times 24$ unit cell metasurface printed on a 500 μm thick PET (polyethylene terephthalate) transparency in which the conductivity varies from 50% silver–50% carbon on the left to 90% silver–10% carbon on the right (the dimensions of the section of the overall array shown are $4.32\mathrm{cm}\times 4.32\mathrm{cm}$. The plastic substrate exhibits $\sim 91\%$ transparency across the measured THz spectral range. The complex refractive index of the PET substrate is measured to be $1.56+\mathrm{i}0.23$ at 0.3 THz. (b) THz transmission spectra for uniform dipole arrays of different conductivities. (c) Schematic showing the THz imaging scheme, with horizontally oriented dipoles as an example. The dipole array is embedded with an image that can only be viewed when probed with the appropriate polarization and THz frequency (top), while no image is obtained with the orthogonal polarization (bottom).

Download Full Size | PDF

We demonstrate the utility of this approach using a randomly generated quick-read (QR) code, which allows for different combinations of conductivity levels in adjacent cells. However, this approach can be used for any type of image or information that can be encoded into gray levels. In Fig. 2(a), we show a design for a QR code that is composed of nine gray levels. We mapped these gray levels to silver–carbon ratios ranging from 50% silver–50% carbon to 90% silver–10% carbon and encoded the spatial information on a unit cell basis using an array of dipoles designed for 0.1 THz. The QR codes were designed using a $72\times 72$ pixel (unit cell) format, with three identical and vertically offset dipoles per unit cell to increase the depth of modulation, corresponding to a sample size of $12\mathrm{cm}\times 12\mathrm{cm}$. In Fig. 2(b), we show the measured THz image, where the transmission magnitude measured for each pixel was mapped back onto the nine gray levels. As noted above, the pattern was imaged using a narrowband THz emitter centered at 0.1 THz, with the incident polarization was aligned along the length of the dipoles. When the incident THz radiation was polarized orthogonal to the dipoles, there was no recognizable image, as suggested in Fig. 1(c). In general, the observed THz image exhibits high fidelity, with only a few errors, typically at the boundaries between large changes in the grayscale level. The pattern was limited to nine gray levels because we found that as the number of levels increased, the number of pixel errors also increased. We defined the pixel error ratio as the ratio of the number of pixels in error to the total number of pixels within each image ($72\times 72$ pixels for each image). Using this definition, the pixel error ratio for Fig. 2(b) was measured to be 2.7% over consecutive measurements.

 

Fig. 2. Nine level QR code embedded using a single dipole-based metasurface. (a) The unit cell design (left) and printed structure (right), in which there are three dipoles in each unit cell to improve the THz response. The length L of the dipole is 1.36 mm and gap G is 300 μm. (b) Design of a randomly generated nine level QR code embedded into a $72\times 72$ unit cell metasurface. (c) The corresponding THz image after mapping the measured image onto the nine gray levels.

Download Full Size | PDF

In order to increase the amount of information that can be stored on the metasurface, we now show that two different images can be embedded into the same pattern using two identically shaped dipoles that are orthogonally oriented, i.e. a cross-shaped resonator. In Fig. 3(a), we show the design and printed version of a unit cell that contains a cross structure, in which each of the cross arms can have different conductivities. Interestingly, even though the conductivity in the overlap region did not correspond to the design parameters of either dipole, we found that better results were achieved with this geometry than with two dipoles that were offset and that did not overlap. In Fig. 3(b), we show two different QR codes designed for the two different polarizations. In order to create the metasurface geometry, we designed cross structures in which the dipole lengths were identical to those used for Fig. 2, with a resonance frequency of 0.1 THz; the cross geometry did not affect the resonance frequencies. Because of the dimensions, only one crossed-dipole structure was printed per unit cell. Here, only six distinct gray levels for each polarization were used, yielding $6^2=36$ levels of information. A larger number of gray levels resulted in more imaging errors, possibly because of the different conductivity in the overlap region of each cross or because of the smaller number of equivalent dipoles in each unit cell. In Fig. 3(c), we show the measured THz images for the two different polarizations. As with Fig. 2, the fidelity is extremely good, with only a small number of pixel errors, often at the boundary between different gray levels. In case of two QR codes mapped onto the cross dipole structure, the pixel error ratio was 3% for each polarization.

 

Fig. 3. Two separate six level QR codes embedded into a metasurface using cross-dipole metasurface. (a) The unit cell design (left) and printed structure (right) corresponding to a pixel (unit cell) on the metasurface. The length of both dipoles is $\mathrm{L}=1.36\mathrm{mm}$ and gaps are $\mathrm{G}=0.3\mathrm{mm}$. (b) Design of a randomly generated two six-level QR codes embedded into the individual arms of the cross-dipole resonator with a $72\times 72$ unit cell metasurface. (c) The corresponding THz images obtained using polarized THz radiation at 0.1 THz after mapping the measured images back onto the corresponding six gray levels.

Download Full Size | PDF

By using dipoles designed for different frequencies, it is possible to further increase the amount of image information embedded within a metasurface. We now use unit cells that have three dipoles of different lengths and different orientations allowing for three color imaging (i.e., each of the three dipoles in the unit cell now corresponds to red, blue, or green channels of information). In order to minimize the issues related to coupling, which would distort the color information, the dipoles were designed such that two of the dipoles were positioned along one axis while the third dipole was oriented along an orthogonal axis, as shown in Fig. 4(a). These dipoles were designed for 0.1, 0.15, and 0.3 THz. Moreover, care was taken to ensure that higher order resonances of the longer dipoles, did not interfere with the fundamental resonances of the shorter dipoles. This was accomplished by orienting the shortest dipole (0.3 THz) orthogonal to the other two dipoles that resonated at lower frequencies but had multiples that might interfere at 0.3 THz. In Fig. 3, we illustrate the implementation of the three-color dipole metasurface, where four distinct gray levels correspond to each color channel, yielding an image with $4^3=64$ possible colors per pixel. To accomplish this, we used three separate QR codes with different gray scale information (details and an additional example are discussed in the Supplement 1). The designed QR code is shown in Fig. 4(b) and the resulting THz image is shown in Fig. 4(c). In general, the three “colors” can be encoded with the same information or totally distinct information. In order to retrieve the information correctly, knowledge of the correct THz frequencies and polarizations would be needed. We found that the pixel error ratio in the individual color images components was 2.5%. However, since some three-dipole pixels had two or three errors, the overall color image had a three-color pixel error ratio 4.7%.

 

Fig. 4. 64 level QR code using a three-color metasurface design. (a) The unit cell design (left) and printed structure (right) corresponding to a pixel (unit cell) on the metasurface. The vertical dipoles have lengths $\mathrm{L}1=1.36\mathrm{mm}$ and $\mathrm{L}3=0.91\mathrm{mm}$ and are designed for resonance at 0.1 and 0.15 THz, respectively. The horizontal dipole has a length $\mathrm{L}2=0.45\mathrm{mm}$ with a designed resonance at 0.3 THz. (b) Design of a 64 color source QR code image generated by overlapping three separate four level QR codes corresponding to the separate RGB color channels. All three QR codes are designed using the $72\times 72$ pixels (unit cell) format. The final image has $4^3=64$ possible colors per pixel. (c) Final THz color image obtained by combining the three separate THz images at 0.1, 0.15, and 0.3 THz.

Download Full Size | PDF

These images can be further hidden by overcoating the image by painting the surface with using conventional spray paints or printing another image over the metasurface using conventional inkjet inks. For the thicknesses required to create a visually opaque coating here, the layer remained largely transparent to THz radiation and did not cause a notable shift in the resonance frequency. However, if thicker coatings are used, changes in the resonance frequencies and amplitude transmission need to be taken into account. Alternatively, the desired pattern can be printed using inks that respond to an external stimulus, such as light, heat or magnetic field, and only appear when simultaneously stimulated via some external means and probed using THz imaging. An example of this idea would involve using graphene or doped semiconductor inks, such that the density of these inks would reveal the image when appropriately excited.

4. CONCLUSION

In summary, we have presented a new approach for hiding images using metasurfaces in which the desired image becomes apparent when probed at the appropriate THz frequency with the appropriate polarization, while ensuring that the fabricated surface looks unpatterned visually. We accomplish this by encoding the relevant pixel information in the conductivity of the dipoles in the unit cell. Using a single dipole, a cross resonator or three different (or differently oriented) dipoles per unit cell, we are able to hide a single image that has 9 gray levels, two separate images using orthogonal polarizations, each with six gray levels yielding 36 unique values, or a color image that has 64 distinct colors. Further refinement in the fabrication and measurement processes is expected to allow for wider variation in the number of allowed image levels. Furthermore, use of alternate types of inks may allow images that can be imaged only when appropriately excited. Finally, we note that this approach can be scaled to higher frequencies, including optical frequencies, using lithographic processes, since the depth of the transmission resonances can be varied by using different plasmonic materials that exhibit different, thereby enabling a wider range of potential applications.