High-resolution photocurrent microscopy using near-field cathodoluminescence of quantum dots

Scanning photocurrent microscopy permits monitoring carrier dynamics under an external excitation in real time, providing rich information on local electronic structures and optical properties in electronics and photonic devices. A laser beam induced current map has been used to investigate lateral inhomogeneity in a variety of semiconductor devices and solar cells, revealing the effect of local impurities and defects on the device performance.^1,2 Using a small fiber probe that can confine optical illumination below the diffraction limit,³ near-field optical scanning microscopy was used to spatially and spectrally resolve photo response of copper indium gallium disulfide (CIGS) solar cell with a lateral resolution of ≈200 nm.⁴ The excess carriers (i.e., electron-hole pairs) in these methods are generated through local photon absorption. Alternatively, an electron beam (typically 1 keV to 30 keV) can directly create the carriers through an ionization process. High-throughput electron beam induced current (EBIC) measurements offer a spatial resolution as high as ≈20 nm both on the top surface and throughout the cross-section of the device, allowing in-situ measurement of the carrier dynamics at the level of individual micro/nanostructures.^5–7 However, the quantitative accuracy of EBIC mapping of low-energy carrier dynamics and thermalization processes may be compromised by the high-energy of the incident electrons. Additionally, direct electron irradiation can cause undesirable artifacts such as charging in poorly conducting materials and damage in organic, carbon-based (e.g., carbon nanotubes and graphene) and biological specimens.^8,9

High-energy electrons interact with a material by losing energy and exciting carriers (electrons and holes) and phonons. Some electrons are scattered back, leaving the sample surface after one or several elastic or inelastic collisions. The energy loss associated with the backscattered electrons depends on the atomic numbers or composition of the materials.¹⁰ The remaining energy of the electron beam is deposited and dissipated in the sample, causing the excitations of electrons and holes, generation of secondary electrons, characteristic X-rays, and photons (i.e., cathodoluminescence) as well as heat.^11,12 Earlier experiments showed that the generation of one electron-hole pair corresponds to the loss of incident electron energy β × E_g, where E_g is the energy band-gap of the material with β ≈ 3 for many inorganic semiconductors and insulators.^13,14 This empirical value, β ≈ 3, accounts for all energy losses mentioned above as well as for energy relaxation of electrons and holes excited well above the band gap.

EBIC signal is generated when the electrons and holes are separated by an internal (or contact) electric field and extracted from the device, while cathodoluminescence (CL) is produced by the radiative recombination of the electron-hole pairs. Recently, CL of a silicon nitride film (ultraviolet photons) was used for optical imaging of latex spheres with a spatial resolution as high as ≈50 nm.¹⁵ This attractive approach can be extended to in-vivo imaging of biological specimens in challenging environments (e.g., liquids).^16,17 In this Letter, we describe a fast, high-resolution photocurrent imaging technique, where a conformal quantum dot film efficiently converts a high energy electron beam into a localized photon source addressable by an electron beam. Q-EBIC (EBIC using quantum dots) results are compared to the corresponding EBIC data using a polycrystalline cadmium telluride (CdTe) solar cell. To assess the luminescence efficiency, CL spectra of quantum dots are compared to those of various semiconductors and phosphors.

Thin-film CdTe solar cells (≈1 cm × 2 cm) used in this work were extracted from a commercial solar module.⁷ An Ohmic contact was made to p-CdTe either by using the metallization remaining on the surface after the extraction process or by evaporating Cu (2 nm) / Au (7 nm) metal contact. A nano-manipulator inside the scanning electron microscope (SEM) with a tungsten probe (100 nm tip radius) wired to an EBIC preamplifier was placed on top of the contact to p-CdTe. The glass substrate coated with n-CdS (cadmium sulfide) / TCO (transparent conducting oxide) layer served as a common contact. To map the efficiency of the carrier collection throughout the entire p-n junction region, cross-sections (5 μm × 40 μm) of the device were prepared using a focused ion beam (FIB). Quantum dots (≈9 nm in diameter) served as CL phosphors, which were deposited as a conformal film on the rough surface of our CdTe solar cells. The homogeneous colloidal cadmium selenide / zinc cadmium sulfide (CdSe/ZnCdS) core/shell quantum dots in toluene showed an emission peak at ≈620 nm with a full width at half maximum (FWHM) of ≈30 nm as determined by either fluorescence spectroscopy or photoluminescence measurement.¹⁸ A dip-coater was used for layer-by-layer assembly of the quantum dots on the CdTe solar cell. We used 0.001 mol/m³ (0.1 M) 1, 3-propanedithiol (PDT) in acetonitrile solution as a cross-linking agent between the CdSe/ZnCdS quantum dot layers.^19,20 All dipping processes were carried out at a rate of 1 mm/s to give a fairly uniform final thickness of the CdSe/ZnCdS quantum dot layer of ≈50 nm, corresponding to ≈6 layers of quantum dots.¹⁸ CL spectra were obtained using a Czerny-Turner spectrometer with a CCD (charged coupled device) camera, where the photons were collected by a diamond-tuned parabolic mirror and dispersed with a grating with a groove spacing of 150 mm⁻¹.²¹ All measurements were performed at a chamber pressure of < 1.3 × 10⁻⁴ Pa (1 × 10⁻⁶ Torr) and the spot size of the electron beam was ≈2 nm. Figure 1 illustrates the schematic of the measurement approaches in this work, where the electron-hole pairs are created by the absorption of photons (Q-EBIC) or through an ionization process by electron beams (EBIC).

A field-emission SEM image shows a uniform coverage of CdSe/ZnCdS quantum dots on a thin metal contact to the CdTe absorber (Figure 2(a)). While the peak-to-valley surface roughness of the CdTe (<500 nm) is larger than the diameter of quantum dot (≈9 nm), the layer-by-layer assembly process provides a highly conformal coating over the large area. The film thickness (≈50 nm) is chosen to be sufficiently thick to absorb all incident electrons for a range of low acceleration voltages while thin enough to enable high-spatial resolution for photocurrent mapping. A set of measured current images of the CdTe solar cell was obtained on the same region for different acceleration voltages, as shown in Figure 2(b)–2(e). The bright contrast seen at many grain boundaries indicates higher carrier collection at grain boundaries than at grain interiors, consistent with our earlier work.⁷ The contact metal (Cu/Au = 2nm/7nm) is sufficiently thick to serve as an electrical contact, but still thin enough to minimize the attenuation of injected electron absorption and photon emission from the CdTe absorber.²² Because the penetration depth of the irradiated electron beam significantly increases with the acceleration voltage, the contribution of the EBIC to the total current becomes more significant than that of Q-EBIC (i.e., photons) at larger acceleration voltages.

Complementary CL spectra of the CdSe/ZnCdS QD coated CdTe specimen were collected at different acceleration voltages (Figure 2(f)—2(i)),¹⁸ illustrating that at an electron energy of 5 keV and below, photon emission (≈620 nm) originated from the quantum dot layer (E_g = 2 eV). Only at higher energies do the electrons penetrate to the underlying CdTe (E_g = 1.5 eV) as evident by the emerging CL peak at ≈810 nm. We note here that quantitative comparison of the CL peak intensities (620 nm vs. 810 nm) is not straightforward. First, the light extraction can be significantly different. Considering photon generation from an incoherent, unpolarized point source and taking the refractive index of CdSe/ZnCdS and CdTe to be 1.7 and 2.9, respectively, the estimated fraction of the photons emitted to vacuum through a planar interface is ≈8.4 % and ≈2 %.²³ Finite-difference time-domain (FDTD) simulations suggest that the fractions increase to ≈16 % for the QD layer and ≈2 % for CdTe when light reflection off a back Cu/Au metal layer is considered.¹⁸ Second, electric fields in the n-CdS/p-CdTe solar cell separate electrons and holes, reducing the corresponding CL intensity at 810 nm. In an ideal solar cell, where the generated electron-hole pairs are separated and collected without any recombination, the CL intensity should vanish. A detailed discussion of CL intensity is provided later in this paper. Nevertheless, the peak of CdTe dominates over the quantum dot emission (Figure 2(h) and 2(i)) with further increase in the electron beam voltage, as the fraction of the total energy deposition in the quantum dot layer diminishes. These results strongly suggest that the obtained Q-EBIC images at low accelerating voltage (<6 keV) are primarily the result of CL photons emitted from the CdSe/ZnCdS quantum dots rather than the direct electron beam excitation of CdTe. In addition, Monte-Carlo simulations of the electron beam absorption in the QD layer support this conclusion (Figure 3(b); discussed below).

To further investigate the distinct local photo response of Q-EBIC in comparison to corresponding EBIC, a systematic study was performed using a FIB cross-sectioned CdTe solar cell. The FIB process results in a smoother surface than the native top surface, thus minimizing possible effects of surface roughness⁷ on light or electron beam injection. An SEM image of a CdS/CdTe cross-section covered with the CdSe/ZnCdS quantum dots (≈50 nm thick layer) for Q-EBIC measurements is shown in Figure 3(a). We performed Monte Carlo simulations²⁴ to estimate the effective size of the light source and the best possible spatial resolution in Q-EBIC (Figure 3(b) and 3(d)). Assuming close-packed CdSe/ZnCdS quantum dots,¹⁸ the estimated range of electron beam absorption extends ≈55 nm at 3 keV (Figure 3(b)). However, over 75 % of the electron energy is absorbed within ≈25 nm thick of the CdSe/ZnCdS quantum dot layer, creating a local photon source. For comparison, Figure 3(d) shows a simulation result of the 3 keV electron beam incident into a bare CdTe layer.

Figure 4 shows a direct comparison between Q-EBIC (Figure 4(a)—4(f)) and EBIC (Figure 4(g)—4(l)) images at different accelerating beam voltages. Overall, the Q-EBIC and EBIC images are quite similar. In both cases, a higher current was observed near the p-n junction, where the internal electric field separates the generated electron-hole pairs. The magnitude of the maximum current increases at larger electron beam voltages in both Q-EBIC and EBIC images. At a high excitation voltage (e.g., 15 keV), the magnitude of the current becomes similar as a consequence of a deep injection of the electron beam. However, a careful inspection of the current distribution of the images indicates a lower spatial resolution in the Q-EBIC than in EBIC at lower accelerating voltage (3 keV), as evident in the vicinity of grain boundaries. Unlike the EBIC, where the electron-hole pairs are generated within the generation bulb determined by the electron beam energy (e.g., ≈20 nm at 3 keV in CdTe), the spatial resolution of the Q-EBIC is determined both by the excitation bulb and by the absorption depth of the photons emitted from the quantum dot layer. Photons with a wavelength of 620 nm travel ≈150 nm in CdTe before the original intensity drops by a factor of 1/e (67 %),²⁵ corresponding to the penetration depth of the electron beam at 7 keV (estimated by a Monte Carlo simulation).²⁴ Thus, the resolution of Q-EBIC image collected at 3 keV parallels the EBIC images collected at 7 keV but a different magnitude of current (i.e., larger number of carrier generation with 7 keV). This limitation of practical resolution in the current experiment also suggests that a possible thickness variation of the QD film will not affect the image if the film thickness is in the range from 20 nm (excitation bulb at 3 keV) to 150 nm (absorption depth of CdTe at 620 nm).

The maximum current measured with Q-EBIC (22 nA) at 3 keV is higher than the current measured with EBIC (17 nA) at the same accelerating voltage. This is perhaps surprising; the band-gap of the quantum dots is greater (E_g = 2 eV) than CdTe (E_g = 1.5 eV), implying less electron-hole pairs created in CdSe/ZnCdS compared to CdTe for a fixed electron beam voltage. The fact that the external quantum efficiency is nevertheless greater for Q-EBIC is reflective of two factors: (1) very efficient photon generation in the CdSe/ZnCdS layer, and (2) Enhanced efficiency of charge collection when carriers are excited by photons emitted by the quantum dots, compared to the excitation by electrons. The latter can be understood as a consequence of the increased absorption depth of photons compared to electrons: generating electron-hole pairs deeper in the interior of the CdTe minimizes surface recombination.

We estimate the internal quantum efficiency of the quantum dot using the measured current (I_{Q − EBIC})in the Q-EBIC experiment. The CL IQE (IQE_{CL; QD})is defined by the number of photons divided by the number of excited electron-hole pairs.

where r_{ehp, QD} is the generation rate of electron-hole pairs in quantum dot layer and f_{CL; QD → CdTe}is the fraction of photon emission from the quantum dot layer to CdTe. For the collection efficiency of CdTe (IQE_CdTe) we use the measured EBIC data. To ensure similar collection efficiency and, respectively, the absorption depth as described above, we compare 3 keV Q-EBIC and 7 keV EBIC data, where the electron-hole pairs were generated within the same volume of the CdTe absorber (i.e., same average extraction efficiency of electrons and holes). A 7 keV electron injected in the CdTe absorber (E_g = 1.5 eV) creates ≈1150 electron-hole pairs by assuming that 3 × E_g electron beam energy is necessary to generate one electron-hole pair¹³ and accounting for the calculated backscattered energy loss of ≈26 %.²⁶ This equals to the maximum current of ≈440 nA under the electron beam current of ≈380 pA. The observed maximum current at 7 keV in EBIC is ≈110 nA, resulting in collection efficiency for CdTe (IQE_CdTe) of ≈25 %. The estimated fraction of photon emission from the quantum dot to the CdTe layer (f_{CL; QD → CdTe}) is ≈90%. This is due to the difference of the refractive indices of the CdTe (n = 2.9), the quantum dot layer (n = 1.7), and the vacuum (n = 1).¹⁸ The generation rate of electron-hole pairs in the quantum dots (r_{ehp, QD}) can again be estimated from the band-gap and the deposited electron beam energy: a 3 keV electron in the CdSe/ZnCdS quantum dot (E_g = 2 eV) layer creates ≈390 electron-hole pairs, suggesting the maximum photocurrent of ≈100 nA at a beam current of ≈260 pA. With these estimates, and the measured current of 22 nA, we find the cathodoluminescence IQE of the quantum dots (IQE_{CL; QD}) is close to ≈100 %. Such high efficiency has been demonstrated for photoluminescence through multiple exciton generation,^19,27,28 yet it has been rarely reported for cathodoluminescence of quantum dots. The accuracy of this evaluation is limited by strong local variations of the electron-hole collection efficiency in the CdTe solar cell.

To independently and reliably measure the internal quantum efficiency of CL, we extensively compare the CL characteristics of the CdSe/ZnCdS quantum dot film with bulk semiconductors, commercial phosphor, and commercial quantum dots.¹⁸ As direct band-gap semiconductors, high purity GaAs (E_g = 1.45 eV) and InP (E_g = 1.35 eV) single crystalline materials have shown high internal quantum efficiencies over 90 %.^29–31 For this comparison, we prepared a thick film (>1 μm) of CdSe/ZnCdS quantum dots by drop-casting onto a bare Si substrate. All spectra were obtained by integrating the CL signal for 1 s, where the entrance slit of the spectrometer was fully open. Figure 5(a) displays the measured CL spectra at 10 keV. The characteristic peaks at 620 nm for QD, 850 nm for GaAs, and 920 nm for InP are matching to their corresponding band-gaps, but the CL intensity of CdSe/ZnCdS QD film is much higher than those of GaAs and InP. This high CL intensity of the QD was observed at different acceleration voltages as shown in the inset to Figure 5(b). The count represents the collected total number of CL photons extracted from each spectrum. Due to the refractive indices of the QD film (n = 1.9), InP (n = 3.3), and GaAs (n = 3.7) at the emission wavelengths (measured by spectroscopic ellipsometery), the fraction of the escaping photons from the materials to the vacuum is different. The estimated CL fraction is 7 % for QD film, 1.7 % for InP, and 1.2 % for GaAs.²³ The estimated backscattered energy loss is ≈28 % for the QD film, ≈27 % for InP, and ≈22 % for GaAs from the Monte Carlo simulations. Figure 5(b) compares the normalized CL signals accounting for the difference in the emission cone (CL fraction), the backscattered energy loss, and the number of generated CL photons. Assuming that IQE is close to 100 % in the high purity GaAs or InP, we calculate IQE of QD film as 102 ± 7 %.¹⁸ This value is consistent with the IQE value determined in the Q-EBIC measurements.

In summary, we presented a new photocurrent imaging approach that combines the high speed and versatility of electron microscopy with optical microscopy by down-converting high energy electrons to local photons through CL in quantum dots. The near-field CL of the CdSe/ZnCdS quantum dot layer is very bright, and the generated photocurrent exceeds the direct EBIC signal within a useable range of parameters. Beyond the immediate application to the local characterization of various types of solar cells and optoelectronic devices, the approach has general advantages for a broad range of microscopies. Sample under study is not exposed to direct electron beam irradiation as the incident beam is fully absorbed in the quantum dots layer conforming to the sample surface. Compared to near-field optical scanning microscopy or other super-resolution optical microscopies, the current approach is high-throughput, can be used for samples with rough surface topography and features the high depth of focus typical for electron microscopy.