Mapping Charge Recombination and the Effect of Point-Defect Insertion in $\mathrm{GaAs}$ Nanowire Heterojunctions

Electronic devices are extremely sensitive to defects in their constituent semiconductors, but locating electronic point defects in bulk semiconductors has previously been impossible. Here we apply scanning transmission electron microscopy (STEM) electron-beam-induced current (EBIC) imaging to map electronic defects in a $\mathrm{GaAs}$ nanowire Schottky diode. Imaging with a nondamaging 80 or 200 kV STEM acceleration potential reveals a minority-carrier diffusion length that decreases near the surface of the hexagonal nanowire, thereby demonstrating that the device’s charge collection efficiency (CCE) is limited by surface defects. Imaging with a 300 keV STEM beam introduces vacancy-interstitial (or Frenkel) defects in the $\mathrm{GaAs}$ that increase carrier recombination and reduce the CCE of the diode. We create, locate, and characterize a single insertion event, determining that a defect inserted 7 nm from the Schottky interface broadly reduces the CCE by $10\mathrm{\%}$ across the entire nanowire device. Variable-energy STEM EBIC imaging thus allows both benign mapping and pinpoint modification of a device’s electron-hole-recombination landscape, enabling controlled experiments that illuminate the impact of both extended (one- and two-dimensional) and point (zero-dimensional) defects on semiconductor device performance.

Figure 1

STEM EBIC imaging of a Au-$\mathrm{GaAs}$ nanowire heterojunction. A low-magnification, 200 kV STEM annular dark-field (ADF) image (a) of a device shows 130 nm diameter $\mathrm{GaAs}$ nanowires and 250-nm-thick, lithographically defined gold contacts supported by a 15-nm-thick silicon nitride membrane. A SEM image [inset in (a)] acquired with ${30}^{\circ }$ stage tilt shows the nanowires as grown, before transfer to the silicon nitride membrane (unlabeled scale bar is 500 nm). When the region indicated in green in (a) is imaged at higher magnification (b), twin boundaries in the $\mathrm{GaAs}$ appear as horizontal lines separated by a few nanometers [Figs. 5–5 show such boundaries more clearly in another device]. An EBIC image (c), acquired simultaneously with (b), reveals the eh separation that occurs near the $\mathrm{Au}$-$\mathrm{GaAs}$ heterojunction. The electrical connections and the locations of the TIA, STEM detectors (ADF, BF), and analog-to-digital converter (ADC) are indicated in a sketch (d).

Figure 2

Mapping eh recombination along the nanowire. A STEM ADF image (a) and EBIC image (b) are acquired simultaneously with a 200 kV accelerating potential. EBIC line profiles (c) are extracted from the center (dashed blue line) and the edge (dashed red line) of the device in (a),(b). The EBIC decay length in the $\mathrm{Au}$ (purple line) and the $\mathrm{GaAs}$ (brown line) measures the radius $R$ of the eh-generation volume $G$ in each material. In the $\mathrm{GaAs}$ far ($>50$ nm) from the interface, the EBIC decay length (green line) measures the minority-carrier diffusion length $L$. All panels are aligned on the same $x$ axis.

Figure 3

Mapping eh recombination across the nanowire. (a) The data of Fig. 2 are rotated ${90}^{\circ }$ to align the Schottky interface with the horizontal axis. Summing the ADF signal from the dashed green ROI in (a) gives a profile (b) approximately proportional to the sample thickness. This profile agrees well with the projected thickness of a geometrically perfect hexagon [dashed red line in (b)]. A slice of the cylindrical eh-generation volume is overlaid on the nanowire cross section (c), with a decay radius $R=10$ nm. The generation volume radii $R$ in $\mathrm{Au}$ and $\mathrm{GaAs}$ and the minority-carrier diffusion length $L$ in the $\mathrm{GaAs}$ are plotted as a function of radial position across the hexagonal nanowire in (d). The $L$ and $R$ measurements shown explicitly in Fig. 2 are highlighted in blue (center) and red (edge) here. All panels are aligned horizontally on the same distance axis.

Figure 4

STEM EBIC at 80, 200, and 300 kV accelerating voltage. Line profiles (a) show the effect of repeated imaging of a device (shown in Fig. 5) at 80 (green), 200 (cyan), and 300 kV (red). Line profiles are extracted from the cyan boxes shown in Fig. 5. Only the 300 kV curves show a significant decrease in the EBIC with repeated imaging. Plotting the profile minima, normalized relative to their initial values, versus dose shows (b) a linear dose effect at 300 kV and insignificant effects at 80 and 200 kV. At 300 kV the maximum EBIC decreases by approximately $3$% per image. Error bars on the 80, 200, and 300 kV data series are determined by setting the reduced ${\chi }^2=1$ for the linear fits. Normalizing the profiles of the 300 kV data series by the minimum value of each (c) shows that only the amplitude of the EBIC line profile changes, not the shape.

Figure 5

Annular dark-field, bright-field, and EBIC imaging before and after irradiation with 300 kV STEM electrons. STEM ADF, bright-field (BF), and STEM EBIC images acquired before (a),(d),(g) and after (b),(e),(h) a dose of $6.0\times {10}^6{e}^{-}/{\mathrm{nm}}^2$ at 300 kV accelerating voltage. The total dose is applied while acquiring the twelve images No. $22$–No. $33$ (Fig. 4) and three alignment images (between No. $21$ and No. $22$). Line profiles are extracted (c),(f),(i) by horizontally averaging data within the blue boxes. A dashed brown line in the line profiles indicates the $\mathrm{Au}$-$\mathrm{GaAs}$ interface. Irradiation produces almost no change in the conventional imaging channels (ADF, BF), but a $44\mathrm{\%}$ decrease in the maximum EBIC, which highlights the advantage of EBIC over conventional imaging for revealing functional properties such as the CCE.

Figure 6

STEM EBIC before and after annealing. A low-magnification STEM ADF image (a) of the device of Figs. 2–5 shows both the heavily irradiated $\mathrm{Au}$-$\mathrm{GaAs}$ heterojunction (red square) and the adjacent control, a heterojunction irradiated less frequently and only at low magnification (cyan arrow). The simultaneously acquired EBIC image (b) shows that the two heterojunctions have EBICs with opposite signs because of their relative orientations in the circuit. The red and cyan rectangles in (b) indicate the sources of the line profiles in the red and cyan plots of (c). After irradiation with 300 kV electrons the magnitude of the irradiated heterojunction’s EBIC is reduced relative to the control. After an anneal the irradiated heterojunction’s EBIC recovers.

Figure 7

Defect insertion and pinpoint localization with STEM EBIC at 300 kV. We record the initial state of an $\mathrm{Au}$-$\mathrm{GaAs}$ nanowire heterojunction with low-dose ($3.0\times {10}^4{e}^{-}$/${\mathrm{nm}}^2$) ADF STEM (a) and STEM EBIC (b) images acquired simultaneously. We then image the region outlined by the dashed box (a) three times (e1),(e2),(e3) with a high dose ($1.8\times {10}^6{e}^{-}$/${\mathrm{nm}}^2$ per image). After the three strip images we acquire a second low-dose EBIC image (c). A difference image (d) shows that the EBIC decreases across the entire nanowire, even though the dose is confined to a narrow region on the left side of the nanowire. Dark-field strip images (e1),(e2),(e3) show no change during irradiation, while the simultaneously acquired EBIC strip images (f1),(f2),(f3) show significantly smaller signals. EBIC difference images (g1),(g2) reveal a sudden drop in the EBIC magnitude within one 0.63 nm pixel, indicating that a defect is inserted during the second strip image at the location indicated by the yellow cross (e2). Enlarged regions [dashed boxes in (g1) and (g2)] of 11 pixels $\times$ 16 pixels ($7$ nm $\times$ 10 nm) demonstrate that both the row and the column of the insertion event can be located precisely. A difference image between the first and third strip images (g3) indicates that, as in (d), the electronic impact of the defect is delocalized. The black-white color scale applies to panels (b),(c),(f), and the blue-yellow color scale applies to panels (d),(g).