Intervalley energy separation in the conduction band of InAs1−x Bi x determined by terahertz emission spectroscopy

InAsBi layers with different bismuth content were grown on InAs substrates by solid source MBE. The amount of bismuth incorporated in the layers was estimated using X-ray diffraction measurements. The relaxation degree of the grown crystalline layers was evaluated using reciprocal space map analysis. The intervalley energy separation in the conduction band of InAsBi was studied by Terahertz Excitation Spectroscopy. It has been found that this separation slightly decreases with increasing Bi content. In the studied samples with Bi content varying from 2.7% to 4.5% the Γ-L separation shifts down to about 0.9 eV.

R ecently, much interest has been attracted by III-V semiconductors containing dilute amounts of bismuth. The incorporation of Bi into the III-V semiconductor lattice results in a rapid reduction of its energy bandgap mainly caused by the valence band anti-crossing effects that are leading to the elevation of the valence band edge. 1) Moreover, spin-orbit splitting of the bismide valence band significantly increases, which reduces parasitic Auger recombination and inter-valence band absorption processes. 2) As yet, most intensively studied were epitaxial layers and quantum wells (QW) of GaAsBi grown on GaAs substrates [3][4][5] ; as a result of these studies, optically and electrically pumped GaAsBi/GaAs QW laser diodes with emission wavelengths up to 1.2 μm [6][7][8] as well as photoconductive terahertz (THz) frequency range components activated by femtosecond 1 μm wavelength laser pulses 9) have been demonstrated. Photoluminescence (PL) at 2.4 μm wavelengths has been observed from GaInAsBi/AlInAs QWs. [10][11][12][13] InAsBi is another dilute bismide alloy that can be epitaxially grown on rather conventional GaSb 14) and InAs 15,16) substrates. InAsBi photodiodes with a spectral response beyond 3.5 μm 17) as well as room temperature PL from InAs 1−x Bi x /In 0.83 Al 0.17 As QWs at 3.1 μm 16) have already been achieved. InAsBi layers have previously been grown by metal-organic vapour phase epitaxy, and the PL results suggested that a bandgap reduction is 55 meV per 1% of Bi. 18) The determination of this parameter is complicated by the uncertain knowledge of the InBi lattice constant value. Moreover, the bandgap reduction rate will be significantly affected by the strain relaxation level in the layer. Therefore, the proposed in the literature bandgap reduction rates of InAsBi still range from 38 meV/%Bi 19) to 60 meV/%Bi. 20) InAsBi could also be a prospective material for fabricating mid-infrared range light sources and photodetectors, replacing the narrow-gap InSb material, known for its light effective mass and strong spin-orbit interaction. The strong bandgap reduction in InAsBi allows one to achieve a similar to the InSb band gap (0.17 eV) inserting as low as 4.5% Bi amount at the same time having significantly lower induced crystal lattice strain, making these structures conventional substrate friendly. The position of the subsidiary valley in the quantum well material is also a known factor limiting performance and electron transition energy in mid-infrared light sources. 21) The separation between Γ and L valleys in the conduction band of InSb is quite low 22) and to have an InAs-like material, which has much larger subsidiary valley separation would be a huge advantage. Therefore, the knowledge of the energy position of these indirect valleys is of a critical importance. In the present contribution, we will describe the measurement of this material parameter in InAsBi by using the THz pulse excitation spectroscopy technique. Three InAsBi samples with different Bi contents were grown on InAs substrates by the MBE and the energy position of the subsidiary conduction valleys in these layers was determined.
InAsBi samples (VIA001, VIA003, VIA012) were grown on InAs (100) substrates using a solid source Veeco GEN Xplor MBE system, equipped with SUMMO group III element sources, Veeco As and Sb valved cracker sources and conventional Dual Filament bismuth source. The substrate temperature was controlled by a thermocouple (TC) and kSA BandIT broadband pyrometry module. The kSA 400 reflection high-energy electron diffraction (RHEED) system was used for in situ surface characterisation.
Each substrate prior to being loaded into the growth chamber has been outgassed at 200°C in the load lock and later annealed at 300°C in the buffer chamber. The native oxide removal was performed at 525°C-530°C temperature according to the pyrometer readings and under ∼1 × 10 5 Torr beam equivalent pressure (BEP) As 2 flux. After that the substrate temperature was decreased to 505°C-515°C for InAs buffer layer growth. The thicknesses of InAs buffer layers in samples VIA001, VIA003, VIA012 were 150 nm, 270 nm, and 120 nm, respectively. The buffer thickness variation from sample to sample is due to different durations of the RHEED intensity oscillation measurements performed as to adjust the growth rate of InAs and III/V ratios prior to the growth of InAsBi layers. InAsBi layers in all three samples were grown at 320°C temperature according to the TC readings. In this case the pyrometer readings were not accessible as the substrate temperature was too low. Thicknesses of grown InAsBi layers in samples VIA001, VIA003, VIA012 are respectively 390 nm, 400 nm, and 500 nm.
Initial conditions for the growth of InAsBi layers were chosen by observing the RHEED pattern change corresponding to the surface reconstruction transition from 2× to 4× at 400°C temperature (pyrometer readings) gradually decreasing the As 2 flux at constant Bi and In fluxes. At lower temperatures the As rich surface corresponding 2 × 4 reconstruction remains stable even without As 2 flux. Therefore, there is no As desorption at temperatures below 400°C and the As 2 flux at 2× to 4× surface reconstruction transition corresponds to the III/V elemental incorporation ratio close to the unity or the group V element sticking coefficient is slightly lower. We supposed that at lower temperatures the sticking coefficients remain the same or will be closer to unity due to lower thermal desorption, making the element V beam slightly excessive. A quite similar III/V ratio for the growth of droplet-free InAsBi layer was reported earlier in Ref. 19.
All the samples were grown at the 0.5 monolayer s −1 growth rate. For the initial sample VIA001 the As 2 /Bi BEP ratio was 12.9 and In/Bi BEP ratio was 15.4. The next InAsBi sample (VIA003) was targeted to incorporate a higher amount of bismuth. For this reason, As 2 /Bi BEP ratio has been decreased to 10.5 keeping the same Bi flux and In/Bi BEP ratio. The structural analysis presented in the next section pointed out that increased amount of bismuth in InAsBi layer caused some strain relaxation and the third sample (VIA012) has been grown using a higher than the initial As 2 /Bi BEP ratio of 13.7 and lower Bi flux (In/Bi BEP ratio was 16.7).
The bismuth content in InAsBi layers was determined from the X-ray diffraction (XRD) (004) rocking curves. XRD curves were measured by Smartlab (Rigaku) diffractometer equipped with Ge (400) 2-bounce monochromator (for Cu Kα 1 X-ray wavelength of 1.5406 Å) and scintillation detector SC-70 (for 2D measurements linear D/tex Ultra detector was used). XRD traces obtained on all three InAsBi samples are presented in Fig. 1. At first, these experimental traces were fitted with the calculations performed assuming that the layers are strained due to small bismuth concentration and thickness not exceeding 400 nm. The lattice parameters and elastic constants of the individual binary constituents in InAsBi used in calculations were a = 6.686 Å, 23) c 11 = 58.5 GPa, c 12 = 30.4 GPa, and c 44 = 27.5 GPa for InBi 24) and a = 6.058 Å, 25) and c 11 = 92.2 GPa, c 12 = 46.5 GPa, and c 44 = 44.4 GPa 24) for InAs, respectively. The flattened top of the InAsBi peak of VIA003 suggests the possible layer relaxation due to higher bismuth concentration. To prove that (115) reciprocal space maps (RSM) of two InAsBi samples with the largest 2θ shifts from the substrate were registered and are presented in Fig. 2. Two strong peaks can be clearly distinguished on these maps: the upper peak can be associated with the InAs substrate, the lower one with the InAsBi layer. It can be seen from this figure that the InAsBi layer with lower Bi content [ Fig. 2(a)] grown on the InAs substrate (VIA001) is strained, whereas InAsBi layers with higher Bi content (VIA003) have a relaxation level of 40%. The composition of all InAsBi layers are as follows: VIA012, VIA001, VIA003 have Bi concentration, 2.7%, 3.6%, and 4.55%, respectively.
The intervalley energy separation in the conduction band of InAsBi epitaxial layers was investigated using the terahertz (THz) excitation spectroscopy (TES) technique. 26) Femtosecond laser pulses illuminating the surface of the semiconductor create in its bulk nonequilibrium current carriers. Because the photoexcited electrons are moving faster from the surface than the holes, this electrical charge spatial separation results in the appearance of a dynamical dipole that radiates into the free space ultrashort pulse of electromagnetic radiation. 25) The amplitude of this pulse is measured by the THz detector as a function of the laser photon energy. When electrons are photoexcited with excess energies higher than the energy separation between the main and subsidiary conduction band valleys, their movement becomes hampered by the scattering to the higher effective mass and lower mobility valleys, and the strength of the dynamical dipole as well as the THz pulse amplitude start to decrease. The photon energies at which this onset begins are then used for determining the inter-valley energy separation in the conduction band of the semiconductor. 26) The experimental set-up was based on an amplified ytterbium-doped potassium gadolinium tungstate (Yb: KGW) laser system (PHAROS, Light Conversion Ltd.) operating at 1030 nm with a pulse duration of 160 fs and a 200 kHz pulse repetition rate. An average power of 6 W from this laser was directed into a cavity-tuned optical parametric amplifier (OPA; ORPHEUS, Light Conversion Ltd.) which generated 140-160 fs duration pulses with a central wavelength tunable from 640 to 2600 nm. In the THz-TES arrangement activated by this laser system, the investigated surface emitters were excited by the output beam from OPA, whereas radiated THz pulses were detected by a photoconducting antenna fabricated from a GaAsBi layer (TERAVIL Ltd.). This THz detector was illuminated by a small part (average power of ∼5 mW) of the Yb:KGW laser beam delayed by different times with respect to the optical beam that was exciting the investigated sample. All experiments were performed at RT. Figure 3 shows a typical THz pulse radiated from InAsBi samples excited by 920 nm wavelength, 7 mW average power optical pulses, and their Fourier spectra. Such bipolar pulses were observed on all three bismide samples over the whole spectral region investigated.    1.4 to 1.6 eV. As the shape of TES dependences around their maxima is most essential for our investigation, this spectral range was additionally measured with a higher resolution. The results of these measurements are presented in Fig. 4(b) for all three investigated InAsBi layers. The TES spectrum of n-type InAs substrate is also shown for comparison. Narrow-gap semiconductors such as InAs or InAsBi radiate THz pulses after photoexcitation of their surface by a femtosecond laser, mainly due to the excited electron and hole spatial separation. In general, the electrons in semiconductors are more mobile than the holes, thus this separation could be caused by their different diffusion rates from the excitation point at the surface-the so-called photo-Dember effect. 27) Moreover, charge separation is facilitated by the optical alignment effect. In cubic semiconductors, electrons excited from the heavy hole valence band have their momenta predominantly perpendicular to the optical field vectors, therefore these electrons propagate in the refracted into the semiconductor optical beam direction. 28) As the momentum relaxation times in InAs and related materials are of the order of several hundreds of femtoseconds, this propagation is nearly ballistic during the electrical dipole formation and THz pulse radiation. 29) The electron and hole separation and radiated THz pulse amplitude are increasing with increasing laser photon energy that leads to the excitation of electrons with larger excess energies and larger group velocities. This increase is effectively stopped when the photoelectron energy exceeds the energy distance to the subsidiary, high effective electron mass and low mobility valleys in the conduction band. Therefore, TES dependences can be effectively used for the determination of the intervalley separation in the conduction bands of various semiconductors. 26,30) When two band Kane law for the energy dispersion in the Γ valley of the conduction band 31) will be used  Figure 5 shows the values of ε ΓL as calculated using formula (2) and the values of the onset of THz amplitude decrease taken from Fig. 4(b). In our calculations we were using the effective masses for electrons and heavy holes-m e and m hh -equal to their values in InAs: 0.026m o and 0.41m o , respectively. 32) On the other hand, we did not succeed in determining the energy bandgap of the bismides from the optical absorption and PL measurements on rather thin and defect-rich samples. Therefore, the value of ε g used for the calculations (also shown on Fig. 5) was based on the Bicontent found from the XRD experiment. The values of the bandgap reduction rate with increasing bismuth content in InAsBi vary from 38 meV/%Bi 19) to 55 meV/%Bi. 18) The data points on Fig. 5 correspond to the calculations using the first set of values, while the bars indicate the variation of the output results when the second set of values is used. The value of the InAs bandgap is sourced from. 33) However, it should be noted that while the same source presents the separation between the Γ and L valleys in InAs, there is no reference confirming whether the presented Γ-L separation value is based on experimental measurements or theoretical calculations. Therefore, we assume that the value of 0.73 eV presented in the source 33) is a theoretical estimation. A similar theoretical estimation of the Γ-L separation in InAs, 0.716 eV, is noted in Ref. 34.
InAsBi samples with various bismuth content were grown by MBE on InAs substrates and characterized by X-ray diffraction measurements. The THz pulse amplitudes emitted from these samples when excited by a femtosecond laser have been measured over a wide range of the laser photon energies. The results of such a measurement were then used for the determination of the inter-valley energy separation in the conduction bands of the investigated materials. It has been found that this separation has decreased rather slightly with increasing Bi content and remains of the order of 0.9 eV for the largest Bi content achieved. This is approximately two times larger than the separation between Γ and L valleys in the conduction band of InSb-semiconductor with a similar energy bandgap as in InAs 0.955 Bi 0.045 . Such a large intervalley separation in InAsBi suggests that these alloys could replace InSb material in InSb-based QW reducing the strain and extending the high electron transition energy limit, which is mainly affected by the position of the L valley in the quantum well rather than by the conduction band offset in quantum well and barrier materials.