Report

DNA-directed nanofabrication of high-performance carbon nanotube field-effect transistors

See allHide authors and affiliations

Science  22 May 2020:
Vol. 368, Issue 6493, pp. 878-881
DOI: 10.1126/science.aaz7435

DNA bricks build nanotube transistors

Semiconducting carbon nanotubes (CNTs) are an attractive platform for field-effect transistors (FETs) because they potentially can outperform silicon as dimensions shrink. Challenges to achieving superior performance include creating highly aligned and dense arrays of nanotubes as well as removing coatings that increase contact resistance. Sun et al. aligned CNTs by wrapping them with single-stranded DNA handles and binding them into DNA origami bricks that formed an array of channels with precise intertube pitches as small as 10.4 nanometers. Zhao et al. then constructed single and multichannel FETs by attaching the arrays to a polymer-templated silicon wafer. After adding metal contacts across the CNTs to fix them to the substrate, they washed away all of the DNA and then deposited electrodes and gate dielectrics. The FETs showed high on-state performance and fast on-off switching.

Science, this issue p. 874, p. 878

Abstract

Biofabricated semiconductor arrays exhibit smaller channel pitches than those created using existing lithographic methods. However, the metal ions within biolattices and the submicrometer dimensions of typical biotemplates result in both poor transport performance and a lack of large-area array uniformity. Using DNA-templated parallel carbon nanotube (CNT) arrays as model systems, we developed a rinsing-after-fixing approach to improve the key transport performance metrics by more than a factor of 10 compared with those of previous biotemplated field-effect transistors. We also used spatially confined placement of assembled CNT arrays within polymethyl methacrylate cavities to demonstrate centimeter-scale alignment. At the interface of high-performance electronics and biomolecular self-assembly, such approaches may enable the production of scalable biotemplated electronics that are sensitive to local biological environments.

In projected high-performance, energy-efficient field-effect transistors (FETs) (1, 2), evenly spaced small-pitch (where pitch refers to the spacing between two adjacent channels within an individual FET) semiconductor channels are often required. Smaller channel pitch leads to higher integration density and on-state performance, but it has the risk of enhanced destructive short-range screening and electrostatic interactions in low-dimensional semiconductors, such as carbon nanotubes (CNTs) (3). Evenly spaced alignment minimizes the channel disorder that affects the switching between on and off states (4). Therefore, although high-density CNT thin films exhibit on-state performance comparable to that of Si FETs (5, 6), degraded gate modulation and increased subthreshold swing (3, 5) are observed because of the disorder in the arrays.

Biomolecules such as DNAs (7, 8) can be used to organize CNTs into prescribed arrays (911). On the basis of the spatially hindered integration of nanotube electronics (SHINE), biofabrication further scales the evenly spaced channel pitch beyond lithographic feasibility (12). However, none of the biotemplated CNT FETs (1214) have exhibited performance comparable to that of those constructed with lithography (15) or thin-film approaches (3, 5, 6, 1618). Additionally, during the surface placement of biotemplated materials, broad orientation distributions (19) prevent their large-scale alignment.

In this work, we showed that small regions of nanometer-precise biomolecular assemblies can be integrated into the large arrays of solid-state high-performance electronics. We used the parallel semiconducting CNT arrays assembled through SHINE as model systems (12). At the FET channel interface, we observed lower on-state performance induced by high concentrations of DNA and metal ions. Using a rinsing-after-fixing approach, we eliminated the contamination without degrading CNT alignment. On the basis of the uniform inter-CNT pitch and clean channel interface, we constructed solid-state multichannel PMOS (p-channel metal-oxide semiconductor) CNT FETs that displayed high on-state performance and fast on-off switching simultaneously. Using lithography-defined polymethyl methacrylate (PMMA) cavities to spatially confine the placement of the CNT-decorated DNA templates, we demonstrated aligned arrays with prescribed geometries over a 0.35-cm2–area substrate. Building high-performance, ultrascaled devices at the biology-electronics interface may enable diverse applications in the post-Si era, such as multiplexed biomolecular sensors (20) and three-dimensional (3D) FETs with nanometer-to-centimeter array scalability.

We assembled DNA-templated CNT arrays using DNA-based SHINE (12). We applied a rinsing-after-fixing approach (Fig. 1A) to remove the DNA templates. Starting from the surface-deposited DNA-templated CNT arrays, both ends of the DNA-templated CNT arrays were first fixed onto a Si wafer with deposited metal bars (first step in Fig. 1A). DNA templates and high-concentration metal salts (1 to 2 M) within the DNA helices were gently removed through sequential rinsing with water and low-concentration H2O2 (second step in Fig. 1A and fig. S5). The inter-CNT pitch and the alignment quality of the assembled CNTs were not degraded during the rinsing process (Fig. 1B and figs. S3 and S4) (21).

Fig. 1 Multichannel CNT FETs with ssDNAs at channel interface.

(A) Design schematic for the rinsing-after-fixing approach. Pink arrows indicate the extension direction of DNA templates and the assembled CNTs. (B) Zoomed-in AFM image along the x and z projection direction for CNT arrays after template removal. White arrows indicate the assembled CNTs. Scale bar, 25 nm. See also figs. S3 and S4 (21). (C) Design schematic for introducing ssDNAs at channel interface and FET fabrication. (D) The Ids-Vgs curves [drain-to-source current density (Ids) versus Vgs plotted in logarithmic at a Vds of −0.5 V] for a multichannel DNA-containing CNT FET before (black line) and after (red line) thermal annealing. See also fig. S7.

To explore the effect of single-stranded DNAs (ssDNAs) at the channel interface, we first fabricated the source and drain electrodes onto the rinsed CNT arrays (Fig. 1C, left). Next, ssDNAs were introduced exclusively into the predefined channel area (first step in Fig. 1C; channel length ~200 nm). Finally, a gate dielectric of HfO2 and a gate electrode of Pd were sequentially fabricated (second and third steps in Fig. 1C and fig. S6).

Out of 19 FETs we constructed, 63% (12 of 19) showed typical gate modulation (on-state current density divided by off-state current density, Ion/Ioff, exceeded 103; fig. S7). The other seven devices exhibited Ion/Ioff < 5, which was caused by the presence of metallic CNTs within the array. At a drain-to-source bias (Vds) of −0.5 V, one typical multichannel DNA-containing CNT FET (Fig. 1D) exhibited a threshold voltage (Vth) of ~−2 V, an Ion of 50 μA/μm (normalized to the inter-CNT pitch) at a gate-to-source bias (Vgs) of −3 V, a subthreshold swing of 146 mV per decade, a peak transconductance (gm) of 23 μS/μm, and an on-state conductance (Gon) of 0.10 mS/μm. Statistics over all of the 12 operational FETs exhibited a Vth distribution of −2 ± 0.10 V, an Ion of 4 to 50 μA/μm, and a subthreshold swing of 164 ± 44 mV per decade (fig. S7A). The transport performance was stable during repeated measurements (fig. S7C).

We annealed the above DNA-containing FETs at 400°C for 30 min under vacuum to thermally decompose ssDNAs (22), and we then recharacterized the transport performance. Compared with the unannealed samples, thermal annealing (Fig. 1D and figs. S7 and S16) slightly shifted the average Vth (~0.35 V, for a Vth of −1.65 ± 0.17 V after annealing) and increased the average subthreshold swing by ~70 mV per decade (subthreshold swing of 230 ± 112 mV per decade after annealing). Other on-state performance metrics, including gm and Gon, as well as FET morphology, did not substantially change after annealing.

To build high-performance CNT FETs from biotemplates, we deposited a composite gate dielectric (Y2O3 and HfO2) into the rinsed channel area instead of introducing ssDNAs (Fig. 2, A and B, and figs. S10 and S11) (21). Of all the FETs constructed, 54% (6 of 11) showed gate modulation (fig. S12). The other 5 of 11 FETs contained at least one metallic CNT within the channel (fig. S15). Using an identical fabrication process, we also constructed another nine operational single-channel DNA-free CNT FETs for comparing transport performance (fig. S8). The single-channel CNT FET (channel length ~200 nm) with the highest on-state performance exhibited an on-state current of 10 μA per CNT (Vds of −0.5 V) at the thermionic limit of subthreshold swing (i.e., 60 mV per decade; Fig. 2C and fig. S9).

Fig. 2 Constructing top-gated high-performance CNT FETs.

(A) Design schematic for the fabrication of top-gated DNA-free FETs. (B) Zoomed-in SEM image along the x and z projection direction for the constructed multichannel CNT FET. Scale bar, 100 nm. See also fig. S11 (21). (C and D) The Ids-Vgs curves (solid lines, plotted in logarithmic scale corresponding to left axis) and gm-Vgs curves (dotted lines, plotted in linear scale corresponding to right axis) for single-channel (C) and multichannel (D) CNT FETs. Blue, red, and black colors in (C) and (D) represent a Vds of −0.8, −0.5, and −0.1 V, respectively. Gray arrows indicate the corresponding axes. See also figs. S9 and S12. (E) Benchmarking of the current multichannel CNT FET in (D) with other reports of high-performance CNT FETs. Device performances from previous publications (3, 5, 1618, 2327) are obtained at a Vds of −0.5 V and channel lengths ranging from 100 to 500 nm. See also figs. S17 and S18.

At a Vds of −0.5 V, the multichannel DNA-free CNT FET (channel length ~200 nm, inter-CNT pitch of 24 nm) with the highest on-state performance (Fig. 2D and fig. S13) exhibited a Vth of −0.26 V, an Ion of 154 μA/μm (at a Vgs of −1.5 V), and a subthreshold swing of 100 mV per decade. The gm and Gon values were 0.37 and 0.31 mS/μm, respectively. The noise in the gm-Vgs curves may originate from thermal noise, or disorder and scattering within the composite gate construct. The on-state current further increased to ~250 μA/μm, alongside a gm of 0.45 mS/μm and a subthreshold swing of 110 mV per decade, at a Vds of −0.8 V.

At a similar channel length and Vds (−0.5 V), we benchmarked the transport performance (gm and subthreshold swing) against that of conventional thin-film FETs using chemical vapor deposition (CVD)–grown or polymer-wrapped CNTs (3, 5, 1618, 2327) (Fig. 2E and figs. S17 and S18). Both high on-state performance (a gm of ~0.37 mS/μm) and fast on-off switching (a subthreshold swing of ~100 mV per decade) could be simultaneously achieved within the same solid-state FET, whereas thin-film CNT FETs with a similar subthreshold swing (~100 mV per decade) exhibited a >50% smaller gm.

When the channel length was scaled to 100 nm, we achieved an Ion of 300 μA/μm (at a Vds of −0.5 V and a Vgs of −1.5 V) and a subthreshold swing of 160 mV per decade (fig. S14). Both the Gon and the gm values were thus promoted to 0.6 mS/μm. The DNA-free CNT FETs exhibited comparable Ion to that of thin-film FETs from aligned CVD-grown CNT arrays (28, 29), even at 60% smaller CNT density [~40 CNTs/μm versus >100 CNTs/μm in (28, 29)]. The effective removal of the contaminations, such as DNA and metal ions, and the shorter channel length contributed to the high Ion. Notably, a previous study had fixed CNTs directly with the source and drain electrodes (13). Because contamination could not be fully removed from the electrode contact areas, the on-state performance (gm and Gon) decreased by a factor of 10.

Furthermore, the subthreshold swing difference between the multichannel (average value of 103 mV per decade) and the single-channel CNT FETs (average value of 86 mV per decade in fig. S9) was reduced to 17 mV per decade. Theoretical simulations suggest that, under identical gate constructs, the uneven diameter of CNTs (6) and the alignment disorder (including crossing CNTs) (5) raise the subthreshold swing (4). We observed a wide diameter distribution of the DNA-wrapped CNTs in atomic force microscopy (AFM) images (fig. S2) and transmission electron microscopy images (fig. S1). Hence, the small subthreshold swing difference above indicated that effective gate modulation and evenly spaced CNT alignment were achieved using SHINE (12) (i.e., the absence of crossing or bundling CNTs within the channel area).

Statistics across all the operational multichannel DNA-free FETs exhibited a Vth of −0.32 ± 0.27 V, an Ion of 25 to 154 μA/μm (at a Vds of −0.5 V and a Vgs of −1.5 V), and a subthreshold swing of 103 ± 30 mV per decade. Different amounts of narrow CNTs (i.e., those with diameters <1 nm) within FETs led to the wide distribution of Ion. Because the Schottky barrier and the bandgap increase with narrower CNT diameters, lower CNT conductance is often observed in narrow CNTs than in those with diameters >1.4 nm (30, 31).

When comparing the transport performance differences between DNA-containing and DNA-free FETs (fig. S16), we observed a largely negatively shifted Vth (−2 versus −0.32 V), a higher drain-to-source current density (Ids) at a positive Vgs (mostly 10 to 200 versus 0.1 to 10 nA/μm), and a more than one order of magnitude smaller gm (4 to 50 versus 70 to 370 μS/μm). Thus, high-concentration ssDNAs and metal ions within multichannel FETs deteriorated the transport performance. Thermal annealing did not fully eliminate the adverse effect because of the presence of insoluble and nonsublimable annealing products, such as metal phosphates (22).

When CNT-decorated DNA templates were deposited onto a flat Si wafer, random orientations of DNA templates were formed through unconfined surface rotation. We solved this issue by using 3D polymeric cavities to confine the surface orientation during large-area placement. We first assembled fixed-width CNT arrays (fig. S19) (21) with a prescribed inter-CNT pitch of 16 nm (two CNTs per array). Next, in a typical 500 μm–by–500 μm write-field on the PMMA-coated Si substrate (with >20 write-fields on a 0.35-cm2 substrate), we fabricated densely aligned crenellated parapet-like PMMA cavities (cavity density of ~2 × 107 cavities/cm2; fig. S20). The minimum and the maximum designed widths of an individual cavity along the z direction were 180 and 250 nm, respectively.

After DNA deposition and PMMA liftoff (Fig. 3A), >85% of the initial cavities (~600 cavities were counted) were occupied by DNA templates (Fig. 3B and fig. S21). The measured angular distribution—defined as the difference between the longitudinal axis of the DNA templates and the x direction of the substrate—was 56% within ±1° and 90% within ±7° (Fig. 3C), per scanning electron microscopy (SEM)–based counting of all of the remaining DNA templates within the 600 cavity sites. This value included improvable effects from the fabrication defects of PMMA cavity sites, the variation during DNA placement, and any disturbance from PMMA liftoff. Notably, the angular distribution was still improved compared with previous large-scale placement of DNA-templated materials (19). CNTs were not visible under SEM because they were embedded within the DNA trenches and shielded from the SEM detector by DNA helices.

Fig. 3 Centimeter-scale oriented placement of fixed-width arrays.

(A) Design schematic for the oriented placement of the fixed-width CNT-decorated DNA templates on a Si substrate. From left to right, the panels show fabricating cavities on a spin-coated PMMA layer, depositing CNT-decorated DNA templates onto the PMMA cavities, and liftoff to remove the PMMA layer. (B) From left to right, zoomed-out and zoomed-in optical and SEM images of the aligned structures on the Si wafer after PMMA liftoff. The scale bars in the bottom left, middle, and right images are 10, 1, and 0.5 μm, respectively. The red rectangles indicate the selected areas for zoomed-in views. The yellow arrows in the right panel indicate the aligned arrays. See also fig. S21 (21). (C) The statistics of counts (left, red axis) and the cumulative percentages (right, green axis) for the aligned structures in (B) at each specific orientation. (D) Plot of angular distributions of the aligned arrays versus the lengths of the DNA templates.

Both the lengths of the DNA templates and the aspect ratio of the PMMA cavities affected the angular distribution. Longer DNA templates (with lengths >1 μm) exhibited narrower angular distribution (0° ± 3.4° in Fig. 3D) than those of shorter DNA templates (with lengths <500 nm, 1° ± 11° in Fig. 3D). Additionally, PMMA cavities with a higher length-to-width aspect ratio (i.e., 10 in Fig. 3B and fig. S20) provided better orientation controllability than those with a lower aspect ratio (i.e., 1 in fig. S22). Hence, longer DNA templates, as well as a higher length-to-width aspect ratio of PMMA cavities, were beneficial in improving the angular distribution. Because PMMA cavities were wider than the DNA templates, we observed up to three DNA templates, as well as the offset of DNA templates along the x and z directions, within a few PMMA cavities. Notably, DNA templates did not fully cover the individual PMMA cavities, even for a saturated DNA solution.

Two-dimensional hydrophilic surface patterns, with shape and dimensions identical to those of the DNA structures, could direct the orientation of the deposited DNA structures (32). However, it is difficult to design patterns adaptive to DNA templates with variable lengths. In contrast, effective spatial confinement relies mainly on the lengths of the DNA templates and the aspect ratio of PMMA cavities and is applicable to irregular template lengths. Therefore, the anisotropic biotemplated CNT arrays with uneven lengths could be aligned along the longitudinal direction of the cavities (supplementary text section S4.1 and fig. S23) (21).

To further promote the on-state performance, scaling the inter-CNT pitch into <10 nm may be beneficial. However, at 2-nm inter-CNT pitch, the enhanced electrostatic interactions may affect the on-off switching. Therefore, the correlation between the inter-CNT pitch and performance metrics of CNT FETs needs to be verified. Combined with large-area fabrications through conventional lithography and directed assembly of block copolymers, biomolecular assembly could provide a high-resolution paradigm for programmable electronics over large areas. The hybrid electronic-biological devices may also integrate electrical stimuli and biological inputs and outputs, producing ultrascaled sensors or bioactuators.

Supplementary Materials

science.sciencemag.org/content/368/6493/878/suppl/DC1

Materials and Methods

Supplementary Text

Figs. S1 to S23

References (3337)

References and Notes

  1. See supplementary materials.
Acknowledgments: Funding: W.S., M.Zha., Y.C., K.W., and Z.Zha. acknowledge the National Science Foundation of China (grant nos. 21875003, 21991134, and 61621061) and Peking University for financial support. Y.C., C.Y., and Z.Zhu. acknowledge the National Science Foundation of China (grant nos. 21775128, 21435004, and 21974113) for financial support. J.K.S., J.A.F., and M.Zhe. acknowledge the NIST internal fund. Author contributions: M.Zha. conducted the experiments on CNT assembly and CNT FETs and analyzed the data; Y.C. conducted the experiments on CNT assembly and centimeter-scale placement and analyzed the data; K.W. and Z.Zha. conducted the experiments on CNT assembly and analyzed the data; J.K.S., J.A.F., and M.Zhe. prepared the DNA-wrapped CNTs and analyzed the data; J.T. analyzed the data; Z.Zhu. designed and supervised the study and interpreted the data; W.S. conceived, designed, and supervised the study and interpreted the data; and all authors wrote the manuscript. Competing interests: Two provisional-stage patent applications were submitted by W.S. and M.Zha. (regarding FET construction) and W.S. and Y.C. (regarding large-area alignment). Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper or the supplementary materials.

Stay Connected to Science

Navigate This Article