Laser direct writing and inkjet printing for a sub-2 μm channel length MoS₂ transistor with high-resolution electrodes

A basic configuration of single line-by-line laser processing is accomplished by implementing a precision translation system on a single platform, as shown in figure 1(a). We use an infinity corrected high magnification objective lens (Mitutoyo, numerical aperture  = 0.5, focal length = 2 mm, and magnification = 100×) to create fine patterns (of ∼2.0 μm width) as well as to observe in situ via a camera. Note that in order to overcome serial direct writing by a single probe, a probe array scheme utilizing a digital mirror display array and piezo-scanners has been developed for increased throughput [26]. A positive photoresist (Fujifilm, OiR-906-12, i-line), which is dissolved within the illuminated area during development, is applied as a photo-sensitive material. We also test and apply the negative photoresist (SU-8 2002, MicroChem), whose non-illuminated area dissolved. Three representative feature trends are observed: the width, height, and even the profile of the resists vary, depending on the laser power and scan speed of the stage (i.e. exposure time). The departures are because the Gaussian beam profile of the laser scales with the peak intensity (see online supplementary figure S1 for detail). Note that the laser-scanned area has a crater shape in the case of positive photoresist, while the behavior of negative photoresist is the opposite (i.e. the laser-scanned area is embossed) as shown in figures 1(b) and (c).

Zoom In Zoom Out Reset image size

2.2.1. Capillary with positively laser-patterned geometry

Grooves are generated by using LDW lithography and positive photoresist. The fabricated patterns are filled with inkjetted Ag ink (CCI-300, Cabot Corp., material properties in online supplementary table S1) to form the S/D electrodes. Capillarity is the driving force for the filling process (figure 2(a)), which exhibits distinct variations, as depicted in figures 2(b)–(d). Therefore, we need to understand the relationship between capillarity and the geometry of the channel. As the patterned channel is open at the top, the capillary pressure is given by [27, 28]:

$$\begin{eqnarray}&&{P}_{{\rm{c}}\_{\rm{o}}{\rm{p}}{\rm{e}}{\rm{n}}}=\gamma \left(\displaystyle \frac{{\rm{\cos }}\,{\alpha }_{{\rm{b}}{\rm{o}}{\rm{t}}{\rm{t}}{\rm{o}}{\rm{m}}}}{d}+\displaystyle \frac{{\rm{\cos }}\,{\alpha }_{{\rm{l}}{\rm{e}}{\rm{f}}{\rm{t}}}+\,{\rm{\cos }}\,{\alpha }_{{\rm{r}}{\rm{i}}{\rm{g}}{\rm{h}}{\rm{t}}}}{w}\right)\end{eqnarray} \tag{ 1 }$$

where γ is the surface tension of the filled liquid, α is the contact angle of the liquid, d is the depth of a channel, and w is the width. The volume flow rate (Q) is expressed in open channels [27, 28] by:

$$\begin{eqnarray}&&Q=\displaystyle \frac{1}{\eta }\displaystyle \frac{{\rm{\Delta }}P}{{R}_{{\rm{f}}}}\end{eqnarray} \tag{ 2 }$$

where η is the viscosity of the ink, ΔP is the pressure difference across the ink front and the ambient air, and R_f is the flow resistance. Because of the complexity of the velocity profile in rectangular ducts, an approximate solution can be constructed through a Fourier series expression. By retaining six terms, the deviation of this prediction from the exact solution is within ∼0.05%. Thus, R_f is given by [27, 28]:

$$\begin{eqnarray}&&\,{R}_{{\rm{f}}}=\displaystyle \frac{12L}{{R}_{{\rm{h}}}^{2}A}(1-1.3553a+1.9467{a}^{2}-1.7012{a}^{3}\,\\ &&\,\,+0.9564{a}^{4}-0.2537{a}^{5})\end{eqnarray} \tag{ 3 }$$

where L is the filled length of channel and A is the cross-sectional area of the channel. R_h is the hydraulic radius of the channel and defined as R_h = 2A/P (P is the wetted perimeter of the cross-section). Moreover, a is the aspect ratio defined as height (h)/width (w) (if h ≤ w) within the range (0 ≤ a ≤ 1).

Zoom In Zoom Out Reset image size

To boost the capillary flow, Q, ΔP should be increased (equation (2)) via reducing w (equation (1)). We note that increasing h is subject to limitations since the thickness of the resist is regulated by the speed of a spin coater and the physicochemical properties of the photoresist, especially the viscosity. Photoresist of ∼1.0 μm thickness is formed by spin coating at 2000 rpm. When a lower speed is used to form a thicker channel, the thickness uniformity worsens. Figures 2(b)–(d) show that decreasing w < 2.0 μm strengthens the capillary force and flow. A similar trend is observed in an additional experiment using a channel with continuously broadening width (online supplementary figure S2). Most of the capillary force and flow into the channel develop in the initial stages, and then diminish due to the increasing R_f (equation (3)) that balances the capillary force. An experiment aiming at filling adjacent channels with Ag ink, separated by variable distance, is also carried out (online supplementary figure S3). After the inkjet printing, the filled ink is sintered at 180 °C for 30 min at a hot plate for improving electrical conductivity and connections.

2.2.2. Self-alignment through negatively laser-patterned geometry

Based on the aforementioned conditions (laser power of 5.0 μW and scan speed of 5 μm s⁻¹ in (section 2.1) irradiated on resist spin coated at 3000 rpm for 60 s) a barrier of ∼2.0 μm thickness is built as shown in figure 3(a). Ag ink is then printed across the barrier as shown in figure 3(b). The jetted Ag ink line breaks since the wall height (∼2.0 μm) exceeds the printed ink thickness (∼430 nm). The ensuing sintering process (at 180 °C for 30 min at a hot plate) further facilitates the Ag line breakup (figure 3(b)) as the differential thermal expansion and the ink shrinkage weaken the line structure. We note that thermal expansion coefficients of Ag and SU-8 are 19.68 ppm °C⁻¹ and 52.0 ppm °C⁻¹, respectively.

Zoom In Zoom Out Reset image size

Through the laser-induced fine patterns (using either positive or negative photoresists) in combination with inkjet printing, multilayered MoS₂ transistors are fabricated. During the LDW process, the irradiated laser does not damage MoS₂ because the applied laser power (5 μW) is very small compared to the ablation threshold of MoS₂ (online supplementary figure S4). The capillarity induced flow for the positive photoresist and the self-alignment for the negative photoresist allow the one-step formation of the S/D electrodes, as shown in online supplementary figures S4(a) and (b); figures 4(a) and (b) show the representative MoS₂ TFTs fabricated by these procedures. The electrical characteristics of the fabricated devices are examined. The log- (left) and the linear-scale (right) I_D–V_G curves of the representative MoS₂ TFT with the S/D electrode formed by capillarity driven filling (V_D = 0.1 V in (figure 5(a)) are plotted. The device has a channel length of 2 μm and a width of 4 μm. Note that the thickness of the MoS₂ flake is around 55 nm. At room temperature and in air, the as-fabricated MoS₂ TFT exhibits acceptable switching behavior (I_on/I_off of ∼10⁴), an SS of ∼2.54 V/decade, a linear field effect mobility (μ_{eff_lin}) of ∼2.87 cm² V⁻¹ s⁻¹, typical n-type behavior with negative threshold voltage (V_TH), and a normally ON-state for the device at low V_D (= 0.1 V). As shown in figure 5(b), the I_D–V_D in a range from 0 to 8 V for the selected V_G (from −20 to +20 V in 10 V increments) is plotted. In the inset of figure 5(b), a non-linear output curve in a very low V_D regime (from 0 to 2.0 V) is observed due to the Schottky barrier (Φ_B) between the MoS₂ and inkjet printed Ag electrodes. MoS₂ TFTs with a channel length of 2 μm and width of 2 μm are also fabricated by the self-alignment method. In figure 6, the electrical performance of representative device exhibits similar characteristics to the previously examined device. At low V_D (= 0.1 V), I_on/I_off of ∼10³ and an SS of ∼3.72 V/decade are obtained and μ_{eff_lin} of ∼10.18 cm² V⁻¹ s⁻¹ is extracted by the MOSFET square-law model (figure 6(a)). Moreover, from the I_D–V_D in figure 6(b), non-linear output characteristics are again observed (see also figure 6(b)).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

MoS₂ TFTs could have large variation in their performances, depending on the structure, process quality, and uncertainty of the geometry of MoS₂ flakes caused by the nature of exfoliation; our extracted μ_eff of both devices is comparable with the low-end of existing vacuum-processed devices (the range: 0.2 to 14 cm² V⁻¹ s⁻¹), however, μ_eff is relatively lower than other vacuum-processed MoS₂ TFTs (the range: 30 to 60 cm² V⁻¹ s⁻¹), as shown in online supplementary table S2. One of the possibilities of changing the electrical behavior is the Φ_B or large contact resistance (R_c) [29–31]. The plots in figures 5(b) and 6(b) confirm that the Φ_B is formed at the contact between the inkjetted Ag electrodes (as a metal) and the MoS₂ flakes (as a semiconductor). Furthermore, the solution-based ink could make an unintended interfacial layer and barrier at the contact due to gas pockets, fully uncapped particles, oxidized particles, or undissolved materials [32]. The undesirable contacts can conceal the inherent performance of the devices. Therefore, to estimate the potential properly, we need to exclude the contact effect. Y-function method (YFM) helps us extract the maximum available mobility (μ₀) and R_c in the device [33–37]. Through the YFM, the extracted μ₀ at a low V_D of 0.1 V are 8.91 cm² V⁻¹ s⁻¹ (improved from 2.87 cm² V⁻¹ s⁻¹, 67.77%) for the device fabricated by capillary force and 22.71 cm² V⁻¹ s⁻¹ (improved from 10.18 cm² V⁻¹ s⁻¹, 55.17%) for the device formed by self-alignment (figures online supplementary S5(a) and (b)). These considerable discrepancies between μ_{eff_lin} and μ₀ originate from the contact effect because the extracted μ₀ is not affected by the contact factor. Also, on the I_D–V_D curves, the non-linear trends arising from the contact barriers are presented clearly at low V_D (figures 5(b) and 6(b)). To further investigate the contact effect, in figure online supplementary S5(c), the mobility attenuator factor (θ = θ₀ (the channel interface) +θ* (the contact)) is considered and through the extracted θ, the maximum value of R_c (R_{c_max}) can be obtained when we assume that the effect of θ₀ is very small (≈0). Note that the calculated values of R_{c_max} are 82.3 kΩ (for capillary) and 64.7 kΩ (for self-alignment). The values of R_{c_max} are 2–3 times larger than well-designed other devices: 17.0 kΩ [21], 27 kΩ [16], 23.4 kΩ [38], 28.9 kΩ [31]. This is because (1) nanoparticle based inks yield reduced properties in comparison with their bulk counterparts (e.g. conductivity: ∼70% of bulk), (2) a thermally sintered Ag electrode shows a somewhat higher work function (4.27–4.72 eV depending on the sintering temperature) compared with that of a vacuum-deposited Ag (4.2 eV), which we expected [39], (3) a slightly oxidized surface causes unintended contact properties [32]. Furthermore, R_{c_max} (82.3 kΩ) formed by capillary forces is larger than R_{c_max} (64.7 kΩ) produced by self-alignment because the self-aligned Ag electrodes are much wider (∼50 μm) and have a larger injecting electrode area (more than 10 times) than the electrodes created by capillarity. Also, the dominant contact resistance contributes to the mobility, especially the linear mobility; μ_{eff_lin} (∼10.18 cm² V⁻¹ s⁻¹, for self-alignment) is much higher than μ_{eff_lin} (∼2.87 cm² V⁻¹ s⁻¹) of the device with the electrodes formed by capillarity. In addition, it should be noted that interface quality is also an important factor affecting the electrical performance of the devices [29, 40]. Through high-k insulators (e.g. HfO, Al₂O₃, ZrO₂, PMMA) phonon scattering can be suppressed and the interface quality can be improved [21, 40–42]. For the aforementioned reasons, MoS₂ TFTs with inkjet printed Ag electrodes exhibit lower electrical performance compared to previous MoS₂ TFTs fabricated by conventional vacuum processes and well-designed contacts [16, 21, 29–31, 38]. Despite its shortcomings, the printing process still draws strong attention because of potentially low cost, low process temperature, compatibility with flexible substrates, and ability to deposit unique features. Moreover, the printing process performance has improved steadily with the development of printable materials.

The morphologies of printed electrodes formed by capillarity and self-alignment are measured through a 3D confocal laser microscope (Olympus, LEXT OLS4000).

100 nm-thick SiO₂ was thermally grown on a heavily doped p-type Si substrate (resistivity < 5 × 10⁻³ Ω cm) as a gate dielectric layer. Then, MoS₂ flakes were mechanically transferred from bulk MoS₂ crystals (SPI Supplies, USA) on the deposited dielectric layer. After that, the positive (OiR-906-12, Fujifilm) or negative (SU-8 2002, MicroChem) photoresist was formed by a spin coating process at 2000 rpm for the positive photoresist and at 3000 rpm for the negative photoresist. Note that the formed thicknesses of the positive and negative photoresist were ∼1.0 μm and ∼1.5 μm, respectively. The LDW lithography through a pulsed laser (A Spectra Physics, Navigator) with 40 ns pulse duration at 20 kHz, (200 kHz maximum frequency) and 355 nm wavelength allowed the definition of concave patterns (positive) as the channels and convex patterns (negative) as the walls. For the S/D formation, nanoparticle type Ag ink (particle size of ∼70 nm, CCI-300 from Carbot Corp.) filled in a 10 picoliter cartridge having a nozzle size of 21 μm was inkjet printed with a drop spacing of 25 μm and translation speed of 7 m s⁻¹ onto the patterned reservoir and barrier to form the capillary force and self-alignment induced S/D, respectively. After printing at room temperature, thus enhancing the filling and separating of the ink, the S/D was sintered at 180 °C for 30 min on a hot plate under atmospheric environment. All the inkjet printing processes were performed using the DMP 2831 (Dimatix, Fujifilm). After the inkjet printing process, jetted Ag inks were sintered at 180 °C for 30 min in air at a hot plate for reducing the resistance and boosting electrical conductivity.