Femtosecond-Laser-Induced Spin-Polarized Electron Emission from a GaAs Tip

TL;DR

This study demonstrates that circularly-polarized femtosecond laser pulses can induce fast, spin-polarized electron emission from GaAs tips without NEA activation, with potential applications in ultrafast spintronics.

Contribution

It shows for the first time that spin-polarized electron emission can be achieved from GaAs tips using femtosecond laser pulses without NEA activation, with detailed measurements of polarization and emission dynamics.

Findings

Electron polarization up to 13.3% achieved.

Emission process duration of 400 fs matches laser pulse duration.

Electron emission currents ranged from 50 pA to 3 nA.

Abstract

It is shown that focusing circularly-polarized 800 nm light pulses of 100 fs duration on the tips of p-GaAs crystalline shards having no negative electron affinity (NEA) activation results in electron emission that is both fast and spin-polarized. The 400 fs duration of the emission process was determined by pump/probe measurements. The three samples we investigated produced electron polarizations of 13.1(9)%, 13.3(7)%, and 10.4(2)%. Emission currents ranged between 50 pA and 3 nA with a sample bias of -100V and average laser power of 100 mW. The electron emission exhibited linear dichroism and was obtained under moderate vacuum conditions, similar to that of metallic tips. This source of spin-polarized electron pulses is "fast" in the sense that the electron emission process is of comparable duration to the laser pulses that initiate it.

Tables

Table 1. Table 1: Polarization and dichroism results for circularly- and linearly-polarized light incident on either the apex (“tip”) or the bulk (“shank”) of three different shard samples.

Target	Light	$P_e(\%)$	D(%)
	Polarization
#1 Tip	Circular	13.1(9)
#2 Tip	Circular	13.3(7)	4.7(6)
	Linear	0.1(5)	41.3(1.0)
#3 Tip	Circular	10.4(2)	1.8(2)
	Linear	2.6(2.5)	18.5(6)
#1 Shank	Circular	1.7(8.0)	6.4(1.4)
	Linear	1.0(2.1)	23.7(5)
#2 Shank	Circular	3.4(1.6)
	Linear	5.2(1.0)

Equations

$$A=\frac{\chi -1}{\chi +1}\mathrm{and}\chi =\sqrt{\frac{C_T{C}_B^{\prime }}{C_T^{\prime }{C}_B}}.$$

$$D\equiv \frac{R_1-R_2}{R_1+R_2};$$

$$\mathcal{R}(\tau )\equiv \frac{R_{both}(\tau )-\left(R_{pump}(\tau )+R_{probe}(\tau )\right)}{R_{pump}(\tau )+R_{probe}(\tau )},$$

$$R_{both}(\tau )=\int {\text{displaylimits}}_{-\infty }^{\infty }{\left[E_{pump}(t)\pm E_{probe}(t+\tau )\right]}^{2n}\mathrm{d}t.$$

Full text

I Introduction

Sub-picosecond, nanometer-scale, spin-polarized electron sources are currently not available. Such a source is desirable for tests of quantum degeneracy and for ultrafast electron microscopy.Lobastov, Srinivasan, and Zewail (2005); Lougovski and Batelaan (2011); Jones et al. (2016); Ropers and Kasevich (2018) The first reported observation of free electron antibunching remains controversial, as the experimental apparatus could not distinguish between the effects of Coulomb pressure and degeneracy pressure.Kiesel, Renz, and Hasselbach (2002); Kodama, Osakabe, and Tonomura (2011); Baym and Shen (2014) As degeneracy pressure is polarization-dependent, while Coulomb pressure is not, a spin-polarized, sub-picosecond, nm-scale source could resolve the controversy. The best combined spatial and temporal resolution in ultrafast electron microscopes is provided by nanotip sources triggered by femtosecond laser illumination, as photocathodes with a planar geometry are limited in spatial resolution by the size of the laser focus.Zewail (2010); Bormann et al. (2015) Direct measurements of the electron pulse duration in ultrafast electron microscopy have shown that the electron and the illuminating laser pulse durations are of the same order.Hassan et al. (2017) Implementing a spin-polarized source into an ultrafast electron microscope would allow for a novel approach to studying magnetic nanostructures at the fs-scale.Wan et al. (2017)

In this work, we present a fast, localized, spin-polarized source of electrons obtained from a sharp p-\ceGaAs bulk [110] crystal shard illuminated with femtosecond laser light. The size of the emission site is approximately $1\text{\,}\mathrm{\SIUnitSymbolMicro m}$ in scale, and the electron polarization achieved so far is $13\text{\,}\mathrm{\char 37\relax}$. The electron emission was studied using methods similar to those developed to characterize pulsed emission from metallic nanotips. Such sources are currently in broad application to produce temporally short electron pulses in beams with high brightness.Barwick et al. (2007); Hommelhoff et al. (2006); Hommelhoff, Kealhofer, and Kasevich (2006); Ropers et al. (2007) They are referred to as “fast,” meaning that the temporal response of the emission process is comparable to that of the light pulse duration, and their spatial resolution has been shown to be determined by the size of the emitter and not by the laser focus used.

Standard CW polarized electron sources use a planar GaAs photocathode that must be layered with, e.g., \ceCs and \ceO2 to lower the vacuum potential below that of the conduction band. This creates a “negative electron affinity” (NEA) condition that allows electron emission by absorption of a single photon from a CW laser (Fig. 1(a)).

When circularly-polarized light with an energy near the bandgap $\Delta$ of \ceGaAs is used to excite electrons, there is an imbalance in excitation probabilities of the two excited ${}^{2}s_{1/2}$ Zeeman substates (Fig. 1(c)),Pierce and Meier (1976) causing the emitted electrons to be spin-polarized. Such sources are used in a variety of fields, including atomic and molecular,Kessler (1985); Gay (2009) high-energy,Prescott et al. (1978); Androic et al. (2013) and condensed matter physics.Feder (1986); von Bergmann et al. (2004); Giebels, Gollisch, and Feder (2013)

Alternative planar photocathodes with and without NEA have been developed to optimize the spin-polarization of the emitted electrons, to provide short pulse operation, and to enhance source brightness. Back-illuminated NEA strained and unstrained thin photocathodes have produced $2.5\text{\,}\mathrm{ps}$ electron pulses with high brightness.Aulenbacher et al. (2002) There, the electron pulse duration is limited by the slow emission process of diffusion through the material. A strained \ceGaAs-\ceGaAsP superlattice with NEA activation resulted in a $16\text{\,}\mathrm{ps}$ pulse durationKuwahara et al. (2012, 2014, 2016), and was used in a spin-polarized transmission electron microscope. The source was determined to have a degeneracy 2 orders of magnitude lower than the cathode tip used to first study free electron degeneracy,Kiesel, Renz, and Hasselbach (2002) with a source size that was limited by the diffraction limit of the laser focus. A planar \ceGaAs photocathode with a Ag overlayer a few nm thick has functioned as a polarized electron source without NEA activation by utilizing a multiphoton electron emission process.Klaer et al. (2013) Electron yields were increased by employing local field enhancement through plasmonic coupling on the surface of a p-doped \ceGaAs wafer, while the spin-polarization of emitted electrons was largely maintained. Pulsed \sim$$100\text{\,}\mathrm{fs} laser light produced a spin-polarization as high as $21\text{\,}\mathrm{\char 37\relax}$, with a value of \sim$$15\text{\,}\mathrm{\char 37\relax} for illumination at a central wavelength of $800\text{\,}\mathrm{nm}$.

Tips of magnetized iron and cobalt-coated tungsten have been used to produce spin-polarized electrons, although these sources have used only CW lasers to date.Irisawa, Yamada, and Mizoguchi (2009); Niu and Altman (2010) Such magnetized sources have a further limitation in that their spin polarization is not optically reversible, unlike that of of \ceGaAs photocathodes. An array of etched \ceGaAs tips, illuminated with CW laser light for both positive electron affinity (PEA) and NEA surface conditions resulted in a maximum polarization of $37\text{\,}\mathrm{\char 37\relax}$, but the electrons were not pulsed.Kuwahara et al. (2006) Implementation of a tip geometry results in field enhancement at the tip apex, which increases the yield of emitted electrons. While a more robust activation surface of layered Cs and Te has been demonstrated,Bae et al. (2018) a tip geometry, as well as a multiphoton emission process, eliminates the need for NEA activation that is sensitive to vacuum conditions.Pirbhai et al. (2013)

The work reported here focuses on obtaining fast, spin-polarized electrons from a sharp p-\ceGaAs bulk [110] crystal shard, which naturally incorporates optical reversibility. To do this, Ti:Sapph pulsed lasers with wavelengths centered around $800\text{\,}\mathrm{nm}$, the appropriate wavelength for single-photon excitation across the band gap, were used to induce multiphoton emission without requiring that the samples have NEA. Fig. 1(b) illustrates this. The vacuum potential (dashed black line) is modified at the surface by the application of a negative DC bias voltage $V$ and the local laser field (solid black line). A single photon with energy just exceeding the bandgap $\Delta$ can promote an electron from the valence band to the conduction band. Absorption of a second photon can in principle result in emission via tunneling through the vacuum potential (blue arrow). Absorption of one or more additional photons provides sufficient energy for the electron to exceed the additional ionization energy $\phi$ and escape into the vacuum (red arrow). The $800\text{\,}\mathrm{nm}$ central wavelength of our lasers accesses the relative excitation probabilities for circularly-polarized light that make standard NEA \ceGaAs sources produce polarized electrons (Fig. 1(c)).

II Experiment

We used two apparatuses, the first to measure electron polarization and emission dichroism, and the second to study the emission process duration and the emission position dependence. Our first optical setup consisted of a Ti:Sapph oscillator (Griffin, KMLabs) with an output that passed through a collimating lens and a periscope placed prior to polarizing optics. A half-wave plate (HWP) followed by a linear polarizer was used to vary the laser power without changing the direction of its linear polarization. The beam then passed through a quarter-wave plate to switch its polarization from linear to left- or right-handed circular. A final HWP was used to rotate the plane of polarization of linearly-polarized beams. The laser then entered the polarization/dichroism vacuum system through a window (Fig. 2). Just before entering the chamber, the width of the laser pulses was measured to be $75\text{\,}\mathrm{fs}$ with a Swamp Optics Frequency-Resolved Optical Gate (FROG).

The vacuum system, with a nominal base pressure of ${10}^{-7}\text{\,}\mathrm{Torr}$, comprised two sections. A sample chamber contained an off-axis front-surface Au parabolic mirror to change the direction and focusing of the laser to a $20\text{\,}\mathrm{\SIUnitSymbolMicro m}$-FWHM spot size. The \ceGaAs shard was mounted on a 3-axis stage to position it in the laser focus. A channel electron multiplier (CEM) near the sample monitored the electron emission current. We also measured the total emission current from the electrically-isolated sample. Emitted electrons were directed to a compact, cylindrical Mott polarimeter,Clayburn et al. (2016)

comprising two concentric cylindrical electrodes and two CEMs placed symmetrically about the entrance that defined the electron scattering plane. The central gold-plated electrode was biased at $+20\text{\,}\mathrm{kV}$, whereas the outer electrode and the mouths of the CEMs were biased at $+500\text{\,}\mathrm{V}$.

To measure the electron polarization, $P_{e}$, the count rates measured by the top and bottom CEMs ($C_{T}$ and $C_{B}$) were monitored for electrons produced by light pulses that were right-hand circularly-polarized, and then compared with the rates when the light helicity was flipped. The electron polarization, $P_{e}$, is given as $P_{e}=S_{\mathrm{eff}}/{A}$, where

[TABLE]

Here, $S_{\mathrm{eff}}$, the “effective Sherman function,” is the polarimeter’s analyzing power, and the primes indicate the CEM rates for left-handed incident laser light. The advantage of measuring $P_{e}$ this way is that it eliminates first-order instrumental asymmetries.Kessler (1985); Gay (2009)

Measurements of the linear and circular emission dichroism were made to better understand the emission process. The dichroism, calculated using total emission as measured by the CEM proximate to the sample, is

[TABLE]

$R_{1,2}$ is the rate of emission for orthogonal polarizations.

Electron emission from the samples was optimized at the edge of the crystal shard. Sharp tip-like shards were made by shattering crystalline wafers and using an optical microscope to determine the “sharpest” pieces.Prins, Jansen, and van Kempen (1996) When using these, total emission currents between $50\text{\,}\mathrm{pA}$ and $3\text{\,}\mathrm{nA}$ were obtained with an average laser power of \sim$$100\text{\,}\mathrm{mW} and a DC sample bias of $-100\text{\,}\mathrm{V}$.

A second, similar apparatus was used to study emission rates as a function of the shard apex morphology, to measure the dependence of emission rate on laser intensity, and to assess the temporal width of the emission process.Jones et al. (2018) Pulses from a Ti:Sapph oscillator (Spectra Physics Tsunami) were focused to a FWHM of $3.6\text{\,}\mathrm{\SIUnitSymbolMicro m}$. The laser pulse intensity FWHM, $\tau_{laser}$, was measured to be $100\text{\,}\mathrm{fs}$. The laser power delivered to the shard apex was controlled by a Brewster window variable attenuator. Pulsed electron emission was detected by a microchannel plate (MCP) placed close to the shard apex, or by an electrometer connected directly to the sample. Prior to entering the chamber, the primary beam was split into pump and probe components in a balanced Mach-Zehnder interferometer. The delay $\tau$ between pump and probe pulses could be adjusted for values between \pm$$4\text{\,}\mathrm{ps}.

When two temporally-separated light pulses hit the sample, the integrated electron emission can be categorized as either “additive” or “super-additive.” Additive emission means that the integrated signal is the same as the sum of the emission from each pulse individually. Super-additivity occurs when the emission is greater than the sum of that due to the individual beams. Additive emission for $\tau>\tau_{0}$ shows that the emission process does not exceed $\tau_{0}$; if $\tau_{0}\approx\tau_{laser}$, the emission process is “fast” as defined in the Introduction. Superadditivity for $\tau\gg\tau_{laser}$ implies the process is slow, e.g., due to thermally assisted processes.Barwick et al. (2007); Hommelhoff et al. (2006); Kealhofer et al. (2012); Jones et al. (2018)

III Results

We first consider the electron pulse emission process. Electron emission from nanotips, if measured to be both nonlinear and additive for $\tau>\tau_{laser}$, has been shown to be fast.Barwick et al. (2007); Hommelhoff et al. (2006); Hommelhoff, Kealhofer, and Kasevich (2006) Our electron emission current shows non-linearity as a function of intensity. It fits with a power law of $n=5.14(16)$ (Fig. 3(a) inset; blue line). The Keldysh parameter for a solid, $\gamma$, characterizes the emission. For $\gamma\gg{1}$, field emission is dominated by multiphoton processes.Bunkin and Federov ; Krüger et al. (2012) Given our focal spot sizes of $20\text{\,}\mathrm{\SIUnitSymbolMicro m}$ and $3.6\text{\,}\mathrm{\SIUnitSymbolMicro m}$, and an average power that never exceeded $150\text{\,}\mathrm{mW}$, our Keldysh parameter readily satisfied this condition in all our experiments and supports our simple multi-photon model. The fifth order non-linearity indicates a five-photon process. (This result is in excess of the three-photon process illustrated in Fig. 1(b)). Generally speaking, the order of the multiphoton process in a given sample can vary with the details of the emitting surface, its local surface electric field, and the nature of surface states near the emission point. *[See; e.g.; Section3.1of][]Schmidt223

Pump-probe measurements as described above were used to determine if the emission was additive.Barwick et al. (2007); Jones et al. (2018) The additivity ratio is defined as

[TABLE]

where $R_{pump}(\tau)$ and $R_{probe}(\tau)$ are the emission rates from the pump and probe beams separately at each delay, and the rate $R_{both}(\tau)$ was modeled as

[TABLE]

The individual pump and probe field amplitudes were modeled as Gaussians with $E(t)=E_{0}\exp[-({t/\tau_{pulse}})^{2}]$. The best fit to the data (red line in Fig. 3(a)) is obtained for $\tau_{pulse}=$ $160\text{\,}\mathrm{fs}$ ($n=5$). The electron emission process is additive ($\mathcal{R}(\tau)=0$ for \tau>~{}$$400\text{\,}\mathrm{fs}) and is thus shown to be faster than this value. Note that this is not a direct measurement of the electron pulse duration. Nevertheless, fast emission processes have so far indicated short electron pulses.Krüger et al. (2012)

We now turn our attention to electron polarization. Measurements of $P_{e}$ were taken with a $20\text{\,}\mathrm{\SIUnitSymbolMicro m}$-diameter focal spot for two focal positions on the three samples we studied. In the first “tip” position, the focal spot was centered on the shard apex. In the second “shank” position, the focus center was moved about $15\text{\,}\mathrm{\SIUnitSymbolMicro m}$ away from the tip towards the bulk. The results of all measurements of $P_{e}$ and emission dichroism, taken with the $20\text{\,}\mathrm{\SIUnitSymbolMicro m}$ focus, are given in Table 1.

In the “tip” position, with circularly-polarized laser illumination, $P_{e}$ was $13\text{\,}\mathrm{\char 37\relax}$ for samples 1 and 2, and $10\text{\,}\mathrm{\char 37\relax}$ for sample 3. Note that these results are comparable to those of Ref. Klaer et al., 2013. Variations in the local structure or p-doping could be responsible for the differences in $P_{e}$.*[See; e.g.; Section3.1of][]Schmidt223 As expected, when the laser was linearly-polarized, the values of $P_{e}$ were consistent with zero. One exception, which we have yet to understand, was observed with sample 2 in the shank position. We note though that this value of $P_{e}$ is less than half that of the polarization measured at the tip with circularly-polarized light.

Finally, we consider the sample morphology. The electron emission rate was found to depend sensitively on the position of the laser focus at the sample. Fig. 3(b) shows a plot of the emission rate measured in a $20\text{\,}\mathrm{\SIUnitSymbolMicro m}$ square area of a shard apex. The two laser focal spot sizes used in this work are shown relative to the size of the $20\text{\,}\mathrm{\SIUnitSymbolMicro m}$ scale bar in the top micrograph. The brightest emission feature was used to measure the emission dependence on intensity (Fig. 3(a) inset) and to perform the pump/probe measurements.

Non-zero linear emission dichroism (Eq. 2) was observed for the GaAs shards similar to a field emission tip (FET). That is, emission is higher when the light’s linear polarization is parallel to the axis of the tip.Barwick et al. (2007); Hommelhoff et al. (2006) In contrast, emission dichroism is absent for standard planar \ceGaAs sources.Kessler (1985); Gay (2009); Pierce and Meier (1976) Dichroism measurements were taken at both focal positions as well. At the tip of the \ceGaAs, the circular dichroism is small (<$$5\text{\,}\mathrm{\char 37\relax}) and the linear dichroism for tips 1 and 2 are $41\text{\,}\mathrm{\char 37\relax}$ and $19\text{\,}\mathrm{\char 37\relax}$, respectively. Linear dichroism measured for tip 1 drops to $24\text{\,}\mathrm{\char 37\relax}$ at the shank, possibly because there is less of a tip-like structure with which the light interacts. Thus our shard “tips” have emission characteristics similar to those of FETsBarwick et al. (2007); Hommelhoff et al. (2006); Hommelhoff, Kealhofer, and Kasevich (2006); Ropers et al. (2007) in terms of nonlinearity, additivity, polarization and local morphology although it is apparent from Fig. 3(b) that the overall shard morphology is complex.

In summary, we have demonstrated a source that is able to produce fast pulses of polarized electrons from a micrometer-size area. This can, in principle, enable the imaging of a small electron spot on a target to measure spin-dependent effects with fs-scale resolution. The reduced vacuum requirements of this source when compared with NEA \ceGaAs sources make it easier and less costly to operate. Although the observed electron polarization is modest, our results demonstrate that this source follows the selection rules illustrated in Fig. 1(c). Polarization might be increased by having a sharper, more well-defined \ceGaAs tip, or varying the laser wavelength. The parameter space is large and open to future study. Through the use of chemical etching and ion milling, it is possible to shape the tip. An optical parametric amplifier can be used to explore the wavelength-dependence of polarization. Investigation of the effects these parameters have on the total yield and polarization of the emitted electrons is needed.

Acknowledgements.

We thank M. Becker for useful conversations and S. Keramati for taking the electron micrographs shown in Fig. 3(a). This work has been funded by NSF awards PHY-1206067 and 1505794 (TJG), EPS-1430519 (HB and TJG), and PHY-1602755 (HB).