Efficiency enhancement of organic thin-film phototransistors due to photoassisted charge injection

Organic semiconductors offer many advantages for certain types of electronics technology due to their light weight, flexibility, and low cost as well as the tunability of their properties through synthetic organic chemistry.^1–5 The further development of photosensitive electronic devices that utilize these types of materials requires an understanding of the underlying physics of charge transport in organic semiconductors under illumination. Although numerous studies published to date have explored the photodetection capabilities of specific organic transistors for potential applications,^6–11 a detailed study relating the number of photons absorbed by an organic transistor to the charges produced as a result of the illumination, which provides the physical basis for the photon-to-charge-carrier conversion efficiency in a photosensitive device, has yet to appear.

The present study takes its model device as an organic thin-film transistor (TFT) with 2,8-difluoro-5,11-bis(triethylsilylethynyl) anthradithiophene (diF-TES ADT) as the active layer. This compound is a functionalized derivative of anthradithiophene that has been extensively studied for its high charge transport, excellent air stability, and solution processability, making it attractive for applications.^12–15 We have carried out a detailed analysis of the carrier generation in diF-TES ADT as a function of the number of photons absorbed by the semiconducting layer at three wavelengths in the visible spectrum: 473, 532, and 633 nm.

Using a shadow mask, gold electrodes were patterned on a Si/SiO₂ substrate and then functionalized to improve packing of the deposited semiconductor. A dilute solution of diF-TES ADT was then spin-coated on top of the gold contacts. The detailed fabrication procedure of the bottom-gate bottom-contact TFT can be found in the supplementary material along with a schematic of the device (Fig. S1). The device under investigation had a channel length and width of 50 and 800 μm, respectively. We verified the operational behavior of the TFT by measuring the current–voltage output characteristics in the dark. We obtained the transfer characteristics under device saturation at fixed drain–source voltage (V_ds) in the dark and under illumination by focusing light from a 473, 532, or 633 nm laser source to a spot diameter of around 15 μm centered between the source and drain electrodes. Representative I–V transfer curves are shown in the supplementary material (Fig. S2) for the measurement in the dark and illumination at low, medium, and high incident laser power. We measured the average power incident on the channel over a period of two minutes using a power meter (ThorLabs PM100D with S121C sensor). By varying the incident laser power on the transistor at room temperature, we investigated the effects of the optical power on the photocurrent (⁠$ΔIds=Idsλ−Ids$⁠), which is the difference between the drain–source current in the dark (I_ds) and under illumination (⁠$Idsλ$⁠) when the transistor was in the saturation regime (⁠$Vds=Vgs=−40$ V). From the photocurrent, we calculated the total number of photoinduced carriers per second (⁠$ΔN=ΔIds/q$⁠) measured at the drain electrode. See Table I for a summary of these quantities.

We measured the optical absorption coefficient from a thin film of diF-TES ADT deposited on a glass slide, using a spectrometer equipped with an integration sphere; we subtracted the background due to the glass slide from the data. We then calculated the total number of photons absorbed at 473, 532, and 633 nm using the thickness of the semiconductor and the optical absorption coefficient, presented in Fig. 1, after accounting for reflections at the semiconductor–silicon dioxide and silicon dioxide–silicon interfaces in the TFT. We obtained an average value of the film thickness of the organic layer by cleaving the transistor in half and getting cross-sectional images using a scanning electron microscope (SEM); see the supplementary material (Fig. S3). We verified that no other elements were present in the organic transistor using EDX and Raman spectroscopy. The power absorbed (P_abs) as a function of the semiconductor thickness (t_s), incident power (⁠$P°$⁠), and absorption coefficient (α) was derived from first principles after taking into account reflections at the air or silicon boundary is

and if contributions from the reflections at the interfaces are ignored, Beer–Lambert law¹⁶ is recovered. For normal incidence, the reflectivity $Rλ$ at optical frequencies can be calculated from the dielectric constant and $RSiλ$ (or $Rairλ$⁠) denotes a wave reflecting at the silicon (or air)–semiconductor interface. The number of photons absorbed every second can then be calculated from the photon energy at each wavelength.

By varying the power of the incident illumination at 473, 532, and 633 nm, the photocurrent (⁠$ΔIds$⁠) was used to compute the number of photoinduced carriers (⁠$ΔN$⁠), i.e., the number of extra carriers due to the effect of illumination. The dependence of the number of photoinduced carriers, $ΔN$⁠, measured on the number of photons absorbed is shown in Fig. 2. The overlapping symbols in the figure (red square, green circle, and blue triangle) at the three incident wavelengths show that carrier generation in diF-TES ADT upon illumination is excitonic in nature as the same number of photocharges is generated for each photon absorbed regardless of the illumination wavelength when the differences in the absorption coefficient are accounted for. Although we used comparable incident illumination powers in all the three cases, the data points for the red wavelength are clustered near the origin, because even at the highest illumination power available to us the number of photons absorbed is still quite small due to the small absorption coefficient at this wavelength. Attaining the same number of absorbed photons with red light as for green or blue light would have required a much higher illumination power at the red wavelength, since the fraction of incident photons that are absorbed is much smaller for red light than for green or blue, as shown in the absorption coefficient curve in Fig. 1. The values on the abscissa in Fig. 2 take into account the incident power and the absorption coefficient to arrive at the number of photons absorbed by the semiconductor. The efficiency of photoinduction depends on the rate of charge generation due to the intensity of the incident light, charge recombination if the mean free path is short and charge transport when the transistor is biased. It is clear that the number of photoinduced carriers collected at the drain electrode per photon absorbed by the semiconductor exceeded the number of photons absorbed at any of the three wavelengths used. We define η to denote the number of carriers collected at the drain electrode for each photon absorbed regardless of whether the carriers arose from creation of an exciton in the semiconductor upon absorption of a photon or from charge injection from the source electrode. An inverse relationship exists between the number of photons absorbed and absorbed photon-to-charge-carrier conversion efficiency, η; our results show that η could reach up to 1400% at the lowest illumination power for a coplanar TFT with a simple fabrication procedure. The responsivity of the transistor, which is the ratio of the output current to the incident optical power, ranges from approximately 140 mA/W for green light to 7 mA/W for red light, proportional to the optical absorption as implied by Fig. 2. We propose a mechanism based on a model incorporating space-charge limited photocharges and photoenhanced charge injection from the electrodes.

Organic transistors based on diF-TES ADT and gold contacts exhibit p-type transport.^17–20 No reports of the electron mobility for diF-TES ADT exist, and our past attempts to measure the electron mobility in devices made from this material have been unsuccessful, likely because the electron mobility is much smaller than the hole mobility. However, theoretical investigations of the electronic and charge transport properties for a similar material [2,8-difluoro-6,13-bis(triisopropylsilylethynyl) anthradithiophene or diF-TIPS ADT] determined a difference between the HOMO and LUMO levels of 2.44 eV and showed that the hole mobility is at least 100 times that of the electron mobility.²¹ The optical absorption curve for diF-TES ADT shown in Fig. 1 suggests that the HOMO/LUMO gap for this material²² is similar to that of diF-TIPS ADT. The absorption of a photon by the semiconductor creates an exciton, which can then dissociate into a free electron and a hole. The free holes contribute to the drain current while the less-mobile electrons accumulate in the channel²³ because their mean free path is much shorter than that of holes and both are smaller than the length of the channel of the semiconductor. Because diF-TES ADT has only exhibited p-type transport with gold contacts, the electron concentration is extremely low in this material (either due to the nature of the contacts, which more efficiently inject into the HOMO, or because electrons are susceptible to trapping much more than holes²⁴). This causes the applied electric field to become non-uniform, leading to space-charge effects—local charge neutrality no longer holds within the organic semiconductor.²⁵ As a result of this imbalance caused by the large difference between the hole and electron mobility, the net charge in the channel of a p-type OTFT when illuminated will be negative and the magnitude of the electric field is greater near the source electrode than the drain electrode, which is at a lower electric potential. The density of photogenerated electrons increases with the increasing illumination power; the increase is less pronounced at very high incident power due to electrostatic shielding near the source electrode. The maximum number of carriers generated due to illumination per second can then be calculated in the space-charge limited model in organic compounds when traps are present²⁶ to be

where W is the width of the channel, t_s is the semiconductor thickness, and μ is the (hole) carrier mobility. $ε°$ denotes vacuum permittivity and ε_s is the relative permittivity of the organic semiconductor, while k_B is Boltzmann's constant and T is the temperature of the semiconductor. E_B, N_t, and $N°$ are the characteristic energy, total trap density, and effective density of states in the valence band, respectively. The density of states was calculated using the Grünewald method in which the density of states is extracted from the response of the linear regime drain current to the electric field resulting from the gate-source voltage.^27–29 The characteristic energy and total trap density were found from an exponential fit of the resulting density of states vs energy.³⁰ $L′$ is the extent of the effective region near the source electrode, which sees a local electric field increase due to the space-charge and was derived by Mihailetchi et al.,³¹ and is given by

$L′$ is inversely proportional to $G′$⁠, the volumetric rate of generation of excitons in the semiconductor. We used the values given in Table II to calculate the number of generated carriers only due to illumination (⁠$Nsclp-trap$⁠) in the space-charge limited model when traps are present; this is shown in the solid orange curve in Fig. 2. This model is a realistic representation of photogeneration in our samples as charge transport is trap-assisted. The number of photogenerated charges, $Nsclp-trap$⁠, given by Eq. (2) and shown in the figure is due to carrier generation in the channel from photoexcitation and does not account for charge injection when the transistor is operational (i.e., it assumes non-injecting or blocking contacts). The number of charges from the gate field is identical in the dark and illuminated measurements as the voltage biases are the same in both cases, and we are only reporting the difference between them. As expected, the space-charge limited model always produces fewer photogenerated carriers than if each absorbed photon were to generate one mobile carrier for every exciton formed (η = 100%), which is an upper theoretical bound for both conventional and organic semiconductors in the absence of trapping and/or recombination (denoted by dot-dashed black line in Fig. 2).

As can be seen in Fig. 2, the number of photoinduced carriers (⁠$ΔN$⁠) measured was significantly greater than the value predicted by the space-charge limited model (⁠$Nsclp-trap$⁠), arising from exciton generation. The dotted curve is a logarithmic fit to the data based on the photovoltaic effect³² in which the photovoltage induced by charge carriers in the active layer gives rise to a significant increase in current. The increase in the photoinduced carrier density compared to the value predicted by the space-charge limited model is a direct consequence of the lowering of the potential barrier between the gold contact and the organic semiconductor, resulting in more efficient charge injection from the contacts into the channel. In a simple equation form, the number of photoinduced charges (⁠$ΔN$⁠) is related to photogeneration of charges coming from excitons (⁠$Nsclp-trap$⁠) by

The efficacy of additional electrons that accumulate at the source decreases at higher illumination power as seen from the shallower slope of the dotted line in Fig. 2 due to electrostatic shielding. The dependence of the extent $L′$ of the space-charge region where electrons accumulate on the number of photons absorbed can be seen on the right-hand axis of Fig. 3. As the figure shows, the space-charge limited model predicts that the space-charge region becomes narrower as the number of absorbed photons increases and more photogenerated electrons accumulate near the source. In the absence of illumination, the extent of the space-charge region is distributed across the whole channel (⁠$L′→∞$ as $G′→0$⁠).

The decrease in the potential barrier between the metallic contact and the semiconductor upon illumination is further demonstrated by the behavior of the threshold voltage of the transistor, calculated from the transfer curves in the saturation regime at fixed V_ds in the dark and under illumination. In an organic TFT, the threshold voltage is the gate voltage that must be applied to fill traps (such as lattice defects, impurities introduced during fabrication, grain boundaries, or interfacial roughness) at the dielectric interface before free charge carriers can accumulate in the channel.³³ The threshold voltage was obtained from the I–V transfer curves by finding where the best-fit line of the $Id−Vg$ plot crosses the x-axis. From Fig. 3, on the left axis, the change in the threshold voltage (⁠$ΔVT=VTλ−VT$⁠), which is obtained by subtracting the threshold voltage in the dark (V_T) from the threshold voltage under illumination (⁠$VTλ$⁠), becomes more positive as the illumination power increases. The shift of the threshold voltage may originate from the electric field generated by the electrons trapped near the source electrode, and hence, the potential is lowered further at higher power since more electrons are generated. This electron accumulation, however, competes with the recombination process that inherently can occur between the trapped electron and injected holes. Although $ΔId(=Idλ−Id)$ and $ΔVT(=VTλ−VT)$ were extracted using two independent methods, the plots of $ΔId$ and $ΔVT$ as a function of the number of photons absorbed are identical because they result from the same mechanism. The photogenerated electrons decrease the potential barrier between the source and the semiconducting channel, as diagrammatically shown in Fig. 4, and also attract additional holes from the electrodes. The net effect is an enhancement in charge injection from the metallic contact causing an increase in the photoinduced charges. The fact that $ΔN$ and $ΔVT$ depend on illumination power in the same way indicates that although some of the injected holes recombine with trapped electrons, this is not the dominant mechanism.

To further support our findings, we calculated the contact resistance using the gated transmission line method (gated TLM) in which the channel resistance is directly proportional to the channel length for a highly uniform active layer.^34–36 The sum of the channel resistance and contact resistance (i.e., the total resistance) scaled by the width of the channel is plotted as a function of the channel length (L) in Fig. 5 in the dark (black round symbols, dashed) and under illumination at a fixed incident laser power of 532 nm (green cross symbols, dotted). We made measurements of multiple identical thin-film transistors with different channel lengths but the same channel width on a single silicon die made in the same fabrication batch and then used the data to evaluate the channel resistance. The total resistance was calculated using an overdrive voltage (⁠$Voverdrive=Vgs−VT$⁠) of 12.5 V in the linear regime to ensure that consistency between devices and the contact resistance was then obtained by linearly extrapolating to L = 0. The contact resistance is lower under illumination, suggesting that hole injection from the source is further facilitated when the transistor is illuminated compared to that in the dark, which validates our model that the potential barrier between the source electrode and the semiconducting channel is decreased.

We have presented here measurements of the number of carriers induced by illumination in a diF-TES ADT transistor as a function of the number of photons absorbed by the semiconductor. Using a space-charge limited model to predict the number of generated photocharges, we find an excess of carriers over the number expected from carrier generation in the channel under illumination at 633, 532, and 473 nm when differences in the absorption coefficient at different wavelengths are accounted for. In our model, this discrepancy is explained by the enhancement of hole injection when the channel is illuminated due to the accumulation of photogenerated electrons near the source electrode, which lowers the potential barrier and, thus, the electric field, for hole injection from the electrode into the semiconducting channel. We base this explanation on three primary observations: (i) the number of photoinduced carriers is larger than the number of photons absorbed, implying that the excess carriers must come from electrodes; (ii) the threshold voltage shifts to more positive values under illumination, meaning that the electric field due to the presence of electrons at the electrode/semiconductor interface is different when compared to the dark measurement; (iii) the contact resistance between the electrode and the semiconductor is lower under illumination than in the dark due to a lowering of the barrier to injection between the electrode and organic semiconductor. When additional photons are absorbed, this effect is less prominent because the build-up of electrons near the source shields additional photogenerated electrons. Although charge injection becomes less efficient at higher illumination powers, multiple carriers are still collected for every photon absorbed by the semiconductor (η > 100%). These findings suggest that organic TFTs with diF-TES ADT or similar materials as the active layer have significant potential for development as highly sensitive photodetectors or other photoactive elements.

See the supplementary material for the device schematic and more-detailed description of fabrication methods, I–V transfer curves, and cross-sectional SEM image of the organic transistor that support the findings of this study.

The authors are indebted to J. E. Anthony for supplying the diF-TES ADT. This work was supported by the National Science Foundation (NSF) under No. DMR-1708379. O.D.J. also acknowledges support from NSF under Award No. ECCS-1810273. The authors also wish to acknowledge the support of Chapel Hill Analytical and Nanofabrication Laboratory (CHANL) for electron microscopy measurements and Energy Frontier Research Center (EFRC) for absorption spectroscopy measurements.