X-ray detection

Our nanowire device consisted of an axial GaAs nanowire p–n homojunction (shown in Fig. 1). The p and n-type sections were doped with Zn and Sn during gold-catalyzed vapor–liquid–solid growth in a metal organic vapor phase epitaxy (MOVPE) reactor resulting into doping concentrations of ~6 × 10¹⁸ cm⁻³, respectively. The resulting axial p–n junction nanowire was transferred to a SiO₂/Si substrate and contacted via two e-beam lithography and metal evaporation steps (details in Supplementary Note 1). This configuration allowed to simultaneously measure XRF and XBIC while applying an external bias voltage. The device was raster scanned across a hard X-ray nanobeam focused to a spot size of 80 × 90 nm² at the nano-analysis beamline ID16B of the European Synchrotron Radiation Facility (ESRF) located in Grenoble, France²¹. In-operando XRF spectra were acquired at different incident X-ray beam energies around the Ga K-absorption edge, while recording the XBIC signal simultaneously. All experiments were conducted under ambient conditions. The current–voltage characteristics were repeatedly measured to monitor the device operation during the measurements (see Supplementary Fig. 3).

Scanning, high resolution in-operando XRF/XBIC measurements were taken in a selected region around the p–n junction. The nanowire was easily located by the Ga XRF signal (Fig. 2a), while the p–n junction appears bright in the XBIC map (Fig. 2b). The XBIC signal results from the electron–hole pairs created in the de-excitation cascade of hot electrons excited above the conduction band by the impinging X-rays²². If electron–hole pairs are generated in the depletion zone of the p–n junction, the internal electrical field efficiently separates the electron–hole pair and drifts the electrons through the junction toward the n-side and the contacts, where they are measured as a current signal. This enables detecting hard X-rays with extremely high spatial resolution at zero bias. The length of the depletion zone of the p–n junction and the nanowire diameter determine the spatial resolution. The measured XBIC signal is finally determined by the convolution of the detector resolution and the spot size of the focused hard X-ray nanobeam. Nanowire-based hard X-ray detectors were already reported with outstanding performance in terms of spatial resolution and XBIC efficiency compared to their bulk counterparts^23,24,25. In particular, a spatial resolution of 0.51 µm was achieved for InP nanowires²³, much higher than in conventional detectors with pixel sizes of tens of microns²⁶. Yet, in order to reach such a resolution, sophisticated and computationally demanding ptychographic reconstruction techniques²⁷ were required.

Remarkably, our GaAs nanowire X-ray detector shows a direct imaging FWHM of the XBIC signal along the nanowire of only ~200 nm (Fig. 2c). Since the spatial extent of the focused X-ray nanobeam is considerably smaller, it does not contribute much to the width of the measured signal. We include it in the detector resolution and thus report an upper bound for the detector resolution directly obtained from the measurements. Perpendicular to the nanowire axis an even smaller FWHM of ~150 nm was measured, which coincides well with the nanowire diameter (~150 nm). Thus, a resolution well below 100 nm should be achievable by reducing the nanowire diameter and engineering the p–n junction. In this way, direct hard X-ray detection can be established at the nanoscale, which allows for scanning high resolution direct mapping in the hard X-ray regime.

The specific shape of the XBIC signal is given by the size of the depletion zone together with the minority carrier diffusion lengths at both n- and p-doped sides²⁸. Minority carrier diffusion lengths L_i for electrons (L_D,n = 68 nm) and holes (L_D,p = 108 nm) were extracted from the respective exponential tails (Fig. 2c; see also Supplementary Fig. 4); they are given by \(L_i = \sqrt {D_i\tau _i}\), with the diffusion coefficients D_i and the minority carrier lifetimes τ_i of electrons and holes, respectively. Thus, we estimated the respective carrier lifetimes for electrons and holes to be 0.6 and 4.6 ps, respectively, using bulk values for D_i²⁹. These carrier lifetimes are much shorter compared to reported bulk values, which is due to the high recombination rate at the nanowire surface^28,30. Although this reduces the efficiency of our hard X-ray detector, it conversely offers a very high spatial resolution. Still, a charge collection efficiency of ~0.4% can be estimated for our device (see Supplementary Fig. 4 and Supplementary Note 3), which is remarkably high considering that the interaction volume is several orders of magnitude smaller than in conventional pixel X-ray detectors²⁶.

X-ray energy dependent in-operando measurements

Further insights can be gained by scanning the incident X-ray energy across a specific absorption edge. Therefore, both nano-XRF and nano-XBIC were measured as a function of the incident X-ray energy around the Ga K-edge (Fig. 3). While XBIC maps allow to locate the p–n junction and to gain insights into the local electric fields, XRF probes the local material composition. The internal electric fields can be manipulated by applying a bias voltage to the p–n junction; furthermore, the detector charge collection efficiency can be improved by adjusting the applied voltage in reverse direction. The measurements were successively conducted for bias voltages of 0, −1, −2, and −5 V in reverse direction (indicated by negative bias voltage values on the y-axis). For each bias voltage, XRF and XBIC maps were recorded at different excitation X-ray energies around the Ga K-edge in steps of 1 eV, ranging from 10.367 to 10.378 keV. The XRF and XBIC maps displayed in Fig. 3 are shown for three representative incident X-ray energies, respectively, and scaled equally for all voltages and energies.

The measurements reveal several findings: the XBIC signal significantly increases for applied voltages in reverse direction due to the stronger electric field at the p–n junction, reaching its maximum at −2 V; simultaneously, however, the spatial resolution drastically decreases due to the expansion of the depletion region in the p- and the n-type segment. Conversely, increasing the voltage above −2 V, lead to a reduced XBIC signal already hinting at the degradation of the device, as shown below. Note that the nanowire device showed no degradation when it was biased in reverse direction without the X-ray beam (see Supplementary Fig. 2 and Supplementary Note 2). Therefore, both an applied bias and the incident X-ray nanobeam are necessary to induce damage.

For X-ray energies below the Ga K-edge, the secondary processes represented by XRF/XBIC signals, which result from the primary X-ray absorption mechanism, are low (first column). Increasing the incident X-ray energy across the Ga K-edge strongly increases both, the Ga K_α XRF as well as the XBIC signals, as visible in Fig. 3. This tendency indicates that 1 s electrons, excited by X-rays at the Ga K-edge region, contribute effectively to the electrical signal, according to the X-ray absorption process. Remarkably, above the Ga K edge-energy (middle column), the Ga K_α XRF signal strongly decreases with increasing voltage in the n-type region.

XANES analysis along the p-n junction

To assess this signal reduction, the edge region of the XANES spectra of the GaAs nanowire for 0 V in the p- and n-type segments are plotted in Fig. 4a (light blue and light orange dots, respectively). Both the p- and the n-type GaAs nanowire segments show similar edge energies and spectral features at the Ga K-edge that are in excellent agreement with those reported for GaAs^31,32, revealing the GaAs short-range structure at zero bias.

However, the edge region of the Ga K-edge XANES spectrum changes drastically for measurements at higher negative bias voltages. Spectra taken at −5 V from the p-and n-type segments (dark blue and dark orange in Fig. 4a, respectively) of the GaAs nanowire clearly reveal a strong shift of the X-ray absorption edge to higher energies for the n-segment, while the p-segment remains unchanged.

There are several works on Ga K-edge XANES measurements that investigate spectral energy shifts in the X-ray absorption edge^32,33,34. In particular, XANES edges for Ga-based materials with different ligands and coordination states of three, four, and six were reported, and edge energies were found to increase with increasing coordination^33,35. In GaAs, Ga has a tetragonal coordination which gives rise to the peak at at ~10.375 keV. Commonly, an increasing coordination of the Ga atoms is observed during oxidization to the stable β phase of Ga₂O₃, which is composed of equal fractions of tetrahedral (GaO₄) and octahedral (GaO₆) sites of the Ga atoms^36,37. Thermal oxidization of GaAs has been studied in literature extensively yielding mostly β-Ga₂O₃, elemental As, and an As depletion^36,38,39.

Therefore, we measured full XANES spectra without applying an external bias after the collection of high resolution XRF/XBIC maps in the p- and the n-type segments near the p–n junction (Fig. 4b); this allows to compare the observed energy shift of the Ga K-edge with literature data. Both XANES spectra are in excellent agreement with the edge region spectra taken at −5 V (blue and orange dots). Thus, the observed energy shift of the XANES edge is permanent, ruling out any transient effect associated to heat or electric fields. All XANES spectra taken in the p-type segment match the GaAs reference well. Thus, the p-type segment remained structurally unchanged throughout all experiments and still consisted of pristine GaAs. The energy shift in the n-type segment on the other hand, in comparison to β-Ga₂O₃ (mixed tetrahedral and octahedral), α-Ga₂O₃ (distorted octahedral), and Ga(AcAc)₃ (octahedral) refs. ^31,32,33, points astonishingly well to Ga(AcAc)₃ in terms of edge energy and post-edge spectral features, representing pure octahedral coordination. The references for α- and β-Ga₂O₃ have XANES edges at ~2 eV lower energies (see Supplementary Fig. 6 for further reference spectra). These findings suggest a selective oxidization of the n-type segment and find Ga in a purely octahedral GaO₆ coordination site (like in Ga(AcAc)₃ or α-Ga₂O₃).

In short, following the evolution of the Ga K_α XRF and XBIC signals as a function of applied negative bias voltage (Figs. 3 and 4), we conclude that a step by step selective oxidization of the n-type area of the nanowire next to the p-n junction takes place. This effect is hinted by the reduction of Ga K_α-intensity in the n-doped region with increasing negative bias voltage for the excitation X-ray energy of 10.372 keV (middle column of Fig. 3), which is between the Ga K-edges of GaAs and octahedrally coordinated Ga. Above an X-ray energy of 10.376 keV, (i.e., above the K-edge of Ga in the octahedral GaO₆ coordination; right column in Fig. 3), the Ga K_α-signal in the n-type segment increases again due to the selective oxidization.

In strong contrast to these observations on the n-doped segment, no changes of the XANES energy edge were observed for the p-type side. But why is the oxidization limited to the n-type side of the p–n junction? To understand the underlying mechanisms of the oxidization, the energy of the Ga K-edge (determined from the edge region XANES spectra, see Supplementary Fig. 5 and Supplementary Note 4) along the nanowire axis is plotted as a function of the position across the p–n junction (Fig. 5a). The XBIC signal profile taken at 0 V is also displayed to exactly locate the position of the p–n junction and used as the origin of the x-axis (x = 0 μm) for all plots in Fig. 5. The Ga K-edge energy in the p-doped segment of the GaAs nanowire is constant at about 10.372 keV for all applied bias voltages. As mentioned above, the energy of the Ga K-edge also stays almost constant across the p–n junction without applied external bias voltages (0 V, black curve in Fig. 5a), revealing the pristine GaAs material. For applied bias voltages of −1 V (blue curve) and −2 V (green curve), only a slight increase of the Ga K-edge energy is visible on the n-doped side of the p–n junction. For a bias voltage of −5 V, however, the Ga K-edge energy on the n-side drastically increases up to almost 10.376 keV (red curve), as discussed above. The intermediate Ga K-edge energy values in the n-type segment most likely originate from a superposition of the GaAs core part of the nanowire with an oxidized surface layer⁴⁰.