Article https://doi.org/10.

1038/s41467-025-59633-1

Low temperature near-field fingerprint

spectroscopy of 2D electron systems in oxide
heterostructures and beyond
Received: 13 March 2023 Julian Barnett 1, Konstantin G. Wirth1, Richard Hentrich2, Yasin C. Durmaz2,3,
Marc-André Rose4, Felix Gunkel 4 & Thomas Taubner 1
Accepted: 25 April 2025

Confined electron systems, such as 2D electron gases (2DEGs), 2D materials, or

Check for updates topological insulators, show great technological promise but their suscept-
1234567890():,;
1234567890():,;

ibility to defects often results in nanoscale inhomogeneities with unclear ori-

gins. Scattering-type scanning near-field optical microscopy (s-SNOM) is useful
to investigate buried confined electron systems non-destructively with
nanoscale resolution, however, a clear separation of carrier concentration and
mobility was so far impossible in s-SNOM. Here, we predict a previously
inaccessible characteristic “fingerprint” response of the prototypical LaAlO3/
SrTiO3 2DEG, and verify it using a state-of-the-art tunable narrow-band laser in
mid-infrared cryo-s-SNOM at 8 K. Our modeling allows us to separate the
influence of carrier concentration and mobility on fingerprint spectra and to
characterize 2DEG inhomogeneities on the nanoscale. Finally, we model the
surface accumulation layer in doped InAs, to show that our fingerprint spectra
are a universal feature and generally applicable to confined electron systems,
like topological insulators or stacked van-der-Waals materials.

Layered heterostructures of complex oxides have become an the spectral information that is observed in s-SNOM is convoluted in
integral part of novel electronic, spintronic, and magneto-ionic device terms of probing depth and in terms of different physical excitations15,
concepts1–4. Such structures often contain high local densities such as phonons, plasmons, or interband transitions. This imposes
of electronic charge carriers, giving rise to correlation phenomena challenges on the qualitative and quantitative understanding of the
and emerging properties, such as metallic behavior5–7, acquired s-SNOM data, especially in layered structures. For one, the
superconductivity8–10, and magnetism11–13, not found in the adjacent near-field response consists of contributions of multiple layers simul-
bulk materials. This poses new challenges to characterize high carrier taneously, which makes a detailed understanding and modeling of the
concentrations in spatially confined and buried electronic systems, near-field response a crucial requirement. Moreover, this calls for the
especially at cryogenic temperatures. Scattering-type scanning near- identification of dedicated spectral regions that provide characteristic
field optical microscopy (s-SNOM)14,15 is a non-destructive method that information on the layer of interest (“fingerprint” regions), adding
uses strong optical near-fields and exhibits high surface sensitivity, requirements to the accessible frequency range of light sources at a
sub-surface capabilities, and applicability in a broad spectral range. sufficient signal-to-noise ratio.
This makes s-SNOM an ideal candidate to investigate highly-confined The 2D electron gas (2DEG) at the interface between the two
electron systems that exist in the transdimensional regime16 between insulators LaAlO3 (LAO) and SrTiO3 (STO)5 has been studied exten-
purely 2D and bulk 3D, such as van-der-Waals materials (e.g., few-layer sively as a model system for high-concentration correlated electron
graphene17) or oxide heterostructures (e.g., LaAlO3/SrTiO35). However, systems. However, the local formation process and the influence of

1
I. Institute of Physics (IA), RWTH Aachen University, Aachen, Germany. 2attocube systems GmbH, Haar, Germany. 3Department of Physics, Ludwig Max-
imilians University of Munich, Munich, Germany. 4Peter Grünberg Institute (PGI-7) and Jülich-Aachen Research Alliance (JARA-FIT), Forschungszentrum Jülich,
Jülich, Germany. e-mail: [email protected]

Nature Communications | (2025)16:4417 1

Article https://doi.org/10.1038/s41467-025-59633-1

defects on the electronic properties are still not fully understood18–20. allows for the separation of carrier concentration and mobility. This
s-SNOM was shown to be sensitive to the 2DEG in LAO/STO21 and even spectral region was previously inaccessible for s-SNOM on LAO/STO
allow in principle, for the extraction of local electronic properties with due to missing light sources with sufficient signal-to-noise ratio (cf.
nanoscale lateral resolution22–24. Due to the metallic behavior of the Supplementary Information S1), as the overall scattering signal of the
system, the 2DEG mobility increases at low temperatures5, which fingerprint region is very low. Here, we use a newly-developed tunable
should lead to a stronger sensitivity of s-SNOM to the 2DEG properties narrow-band mid-infrared laser to investigate LAO/STO in the 2DEG
at cryogenic temperatures (cryo-s-SNOM) compared to room tem- fingerprint region, using cryo-s-SNOM at 8 K. This allows us to use a
perature. Previous s-SNOM studies were limited to indirectly probing normalization procedure that is highly sensitive to the influence of the
the 2DEG via secondary effects, i.e., damping of the phonon near-field 2DEG (cf. Supplementary Information S2) and can be used to char-
resonance of the STO substrate23 or as a constant background at higher acterize the local electronic properties in detail.
frequencies22,24. During these studies, direct mapping of the local
charge carrier density turned out to be difficult, as the influence of Results
different parameters, e.g., carrier concentration and mobility, can Figure 1a presents a simplified view of the experimental setup (cf.
compensate, resulting in similar near-field spectra24. Methods for details), with the metal-coated AFM tip scanning across
In this work, we predict the existence of a spectral fingerprint the LAO/STO sample while oscillating with tapping frequency Ωtip and
region, where a characteristic scattering response of the 2DEG can be being illuminated with a focused laser beam. The measurements pre-
obtained directly, resulting in a near-field spectroscopic method that sented in this work were recorded with a cryo-neaSCOPE by attocube

Fig. 1 | Investigation of LaAlO3/SrTiO3 (LAO/STO) interfaces using scanning LAO/STO (violet), and LAO/STO with 2DEG (dashed orange) at room temperature.
near-field optical microscopy (s-SNOM) and spectroscopy in the 2D electron Grey-shaded areas indicate spectral regions that have been investigated at room
gas (2DEG) fingerprint region. a Experimental setup showing the LAO/STO sam- temperature before22–24. e Simulation of s2/s2Au at cryogenic temperatures (with
ple (cooled to 8 K) and the metal-coated AFM tip oscillating at tapping frequency increased 2DEG mobility). The shaded areas indicate the spectral range investi-
Ωtip, illuminated by a new fast-tunable narrow-band mid-infrared laser. b Real (solid gated at low temperatures previously22 (grey) and in this work (green). f Zoom-in of
lines) and imaginary part (dashed lines) of the dielectric functions of STO (blue) and Fig. 1d to low scattering amplitudes in the fingerprint region, showing the insulating
LAO (red). Both oxides show phonon resonances in the mid-infrared range. c STO (violet) and conducting (dashed orange) LAO/STO samples. g By referencing the
single crystals (left) were epitaxially covered with 8 uc of LAO. Depending on the conducting to the insulating case (dividing of scattering amplitude), changes
termination of the STO substrate, the interface is either insulating (SrO-terminated) introduced by the 2DEG are highlighted, resulting in a characteristic response with
or conducting (TiO2-terminated), the latter hosting a 2DEG. d Simulation of the Au- three separate maxima (fingerprint spectrum).
normalized near-field amplitude spectra s2/s2Au for STO substrate (blue), insulating

Nature Communications | (2025)16:4417 2

Article https://doi.org/10.1038/s41467-025-59633-1

systems GmbH, with the samples being at a constant temperature of resonance”) slightly below its LO-frequency νLO = 788 cm−1, where
8 K. For illumination, the narrow-band (linewidth <5 cm−1) laser PT277- Re[εSTO] is slightly negative and Im[εSTO] is low. When adding the
XIR by EKSPLA was used, which is based on a picosecond optical ultra-thin LAO on top of STO (violet), the overall shape of the STO
parametric oscillator with difference frequency generation and exhi- near-field resonance persists but is modified by the zero-crossings of
bits a high signal-to-noise ratio in the fingerprint region of the LAO/ Re[εLAO] around 600, 650, and 750 cm−1, resulting in amplitude fea-
STO 2DEG. While the spot size of the focused laser on the sample is tures with derivative line shape. Adding 2DEG charge carriers at the
diffraction-limited to λ/2 (several micrometers in mid-IR), the near- interface (dashed orange), the scattering amplitude of the near-field
fields at the tip apex depend on the tip radius and enable a lateral resonance around 700 cm−1 is reduced by additional damping and
resolution down to the ten-nanometer range. The back-scattered light the high-frequency scattering above 900 cm−1 is slightly increased
is demodulated at higher harmonics nΩtip of the tapping frequency, (barely visible at this scale). Grey-shaded areas indicate spectral
and pseudo-heterodyne interferometric detection allows for separa- regions where the 2DEG was experimentally investigated in previous
tion into near-field scattering amplitude sn and phase ϕn. In s-SNOM, studies. These studies were limited to the influence of the 2DEG on i)
the absolute scattering signal strongly depends on the experimental the phonon near-field resonance of STO (<750 cm−1), where the
setup, e.g., the tip shape, oscillation amplitude, detector sensitivity, or s-SNOM signal is generally high23, or ii) the off-resonance scattering
laser power. To acquire setup-independent results, it is an established response in the spectral window of light sources with a high signal-to-
technique to use normalization to a reference sample, resulting in the noise ratio, such as CO2 lasers (>920 cm−1)22,24. In both regions, the
relative scattering amplitude sn/snref and phase ϕn−ϕnref. general influence of free charge carriers on near-field spectra is only
This near-field scattering amplitude and phase depend on the indirect, either i) as an additional source of damping (increased
frequency-dependent optical properties of the sample, expressed as Im[ε]) or ii) as a constant background (increased high-frequency limit
the dielectric function or permittivity ε(ν). Figure 1b shows the real ε∞, cf. Supplementary Information S3). Thus, a clear separation of
(solid lines) and imaginary (dashed lines) parts of the permittivity of carrier concentration and mobility was not previously possible.
bulk STO (blue) and bulk LAO (red), both of which exhibit phonon
resonances in the mid-IR range. The near-fields of s-SNOM decay Near-field fingerprint spectroscopy in LAO/STO
exponentially with distance to the tip, in the order of several ten Figure 1e shows how the calculated near-field spectrum of the con-
nanometers25–27, thus even layers of a few nanometers thickness can ducting LAO/STO interface (dashed orange) is expected to change at
significantly influence the scattering signal28–30, especially in the vici- low temperatures, where the higher 2DEG mobility leads to a slight
nity of zero-crossings of the materials’ Re[ε]23,31,32. The probing depth shift of the peak to higher frequencies and an increased scattering
also depends on the experimental parameters33, and can thus be amplitude. Zooming-in to frequencies above the phonon near-field
adjusted between several tens and hundreds nanometers. For our resonance, Fig. 1f shows significant differences between the scattering
layered system, this leads to a combined scattering response of the amplitude of the SrO-terminated (without 2DEG, violet) and the TiO2-
STO substrate, the LAO top layer, and the 2DEG in between, dependent terminated LAO/STO interface (with 2DEG, dashed orange). Adding
on their respective permittivities. To isolate the influence of the 2DEG the 2DEG leads to three changes in the spectrum, as indicated by black
layer from the contributions of the LAO covering layer and STO sub- arrows: (I) a shift of the high-frequency flank of the STO phonon near-
strate, a direct comparison of samples with and without 2DEG, field resonance to higher frequencies, (II) an additional spectral feature
respectively, is helpful, ideally showing identical lattice phonon around 775 cm−1, and (III) a general increase in scattering amplitude.
properties. This can be achieved by controlling the termination of the However, the overall s-SNOM signal in the spectral range above
STO substrate (Fig. 1c): SrO-terminated STO surfaces lead to an insu- 750 cm−1 is quite low, with relative scattering amplitudes between 2%
lating interface (without 2DEG) after LAO deposition, while TiO2-ter- and 30% of that of gold (cf. Fig. 1d, e), and thus a high signal-to-noise
minated surfaces lead to a conductive interface (with 2DEG), if a critical ratio of the illumination source is necessary to record useable near-
thickness of at least 4 unit cells (uc) of LAO is deposited epitaxially5,24. field spectra. This is especially true for our proposed normalization
The samples used in this work were fabricated by pulsed laser procedure between samples with and without 2DEG, respectively,
deposition (PLD) of 8 uc (3 nm) LAO on single crystal STO substrates. which is the most sensitive to the 2DEG properties but also the most
All substrates were wet-etched to get pure TiO2 termination, and then susceptible to artefacts from referencing to noise.
for one of them, a single SrO layer was applied by PLD, to invert the Figure 1g showcases the theoretical prediction of this character-
termination (cf. Methods for details)34. These two types of LAO/STO istic 2DEG near-field response, which can be obtained by referencing
samples fabricated in identical conditions from a split STO substrate the sample with 2DEG to the sample without 2DEG, i. e. by using Fig. 1e
enable a comparable measurement of conductive and insulating and dividing the data plotted in dashed orange by that in solid violet.
interface. This allows for a direct investigation of the 2DEG response at The resulting normalized amplitude spectrum shows only changes to
low temperatures by using the non-conducting interface as a reference the spectrum induced by adding the 2DEG charge carriers to the non-
in the s-SNOM measurement without introducing referencing conducting interface, with three distinct peaks that relate to the
artefacts. described changes I-III, introduced in Fig. 1e. As the height, position
To showcase the influence of the different layers, Fig. 1d pre- and shape of these features are very sensitive to changes of the elec-
sents simulations of the gold-normalized near-field scattering tronic properties (cf. Fig. 2), they can be seen as a characteristic fin-
amplitude s2/s2Au for bulk STO (blue), LAO/STO without 2DEG (vio- gerprint near-field spectrum of this 2D electron system, similar to
let), and LAO/STO with 2DEG (dashed orange), at room temperature characteristic vibrational bands of polymers in near-field
(cf. ref. 23). For all simulations and measurements, the demodulation spectroscopy39. We have thus identified a unique fingerprint sig-
order n = 2 is presented here, to get a high signal-to-noise ratio with nature of the LAO/STO 2DEG, with three characteristic peaks in the
sufficient background suppression in the observed spectral range. spectral range of 700 to 900 cm−1. Note that this fingerprint
Simulations were done using the finite dipole model35 for the near- spectrum is constrained to the shown frequency range by the phonon
field calculations, combined with the transfer matrix method to get properties of the two materials and could not be experimentally
the near-field response of an arbitrary layer stack (cf. Methods for investigated in previous publications due to a lack of light sources with
details)36. The 2DEG was modeled as several layers of exponentially sufficient signal-to-noise ratio (cf. Fig. S1 in the Supplementary
decaying carrier density22, in accordance with depth profile Information).
assumptions from literature37,38. The STO substrate alone (blue) To illustrate how these fingerprint spectra of the 2DEG (cf. Fig. 1g)
shows a strongly enhanced scattering amplitude (“near-field are influenced by the electronic properties, Fig. 2 shows theoretical

Nature Communications | (2025)16:4417 3

Article https://doi.org/10.1038/s41467-025-59633-1

Fig. 2 | Fingerprint spectra of the 2DEG at conducting LAO/STO interfaces, 6 × 1013 cm−2 (yellow), with a constant µ = 20 cm²/Vs. The inset shows the maximum
normalized to the non-conducting interface. a Simulation for 2DEG mobilities µ value of peak I) with rising carrier concentration. c Measurement of fingerprint
between 5 cm²/Vs (dark red) and 80 cm²/Vs (yellow), with a constant 2DEG carrier spectra at two different positions of the conducting sample. d Simulation of fin-
concentration n2D = 4 × 1013 cm-2. The inset shows the maximum value of peak I) with gerprint spectra with different simulation parameter sets (cf. Supplementary
rising mobility. b Simulation for different n2D between 2 × 1013 cm−2 (dark red) and Information S3), reproducing the experimental spectra shown in c.

predictions of the changes introduced by a varying 2DEG mobility µ small peak (II) at 790 cm−1 and a broad peak (III) around 850 cm−1 can
(Fig. 2a) and 2DEG sheet carrier concentration n2D (Fig. 2b). Values for be observed, which generally fits well to the theoretical predictions
carrier concentration and mobility were picked in accordance with Hall presented in Fig. 2a/b, especially for peak (I) and (II). The fingerprint
measurements of the sample (cf. Supplementary Information S4), with spectrum recorded at the second position (grey) deviates sig-
experimental values of n2D = 4.3 × 1013 cm−2 and µ = 5.1 cm²/Vs at room nificantly from the first, with a lower intensity in the first (I), a higher
temperature. While the 2DEG mobility can increase to above intensity in the second (II), and a less defined shape in the third peak
104 cm²/Vs upon cooling5, it should be noted that the effective mobility (III). Additionally, the minimum around 775 cm−1 and the maximum
in the mid-infrared spectral range can be significantly lower than those around 850 cm−1 are both shifted to lower frequencies by 5–10 cm−1.
measured in the far-infrared spectral range or in transport measure- Figure 2d presents simulations with different parameter sets, to
ments, due to spectral weight redistribution from polaronic reproduce the measurements from Fig. 2c: set A (black) was achieved
contributions38,40. In Fig. 2a, a rising mobility (transition from dark red with n2D = 6 × 1013 cm−2 and µ = 20 cm2/Vs, while set B (grey) relates to
to yellow) from 5 to 80 cm²/Vs leads to an increase of the relative n2D = 8 × 1013 cm−2 and µ = 10 cm2/Vs. Generally, good agreement
scattering amplitude for the peak around 750 cm−1, while the other two between 2DEG fingerprint spectra and simulations can be achieved,
peaks remain mostly unchanged. Contrary to that, a rise in carrier showing that the theoretical description of the 2DEG with a finite
concentration (transition from dark red to yellow in Fig. 2b) leads to an dipole model, transfer matrix method, exponentially decaying depth
increase of the relative scattering amplitude for all three peaks. Addi- profile, and phonon background represents the physical behavior of
tionally, the scaling behavior is very different, as can be seen from the the system well. The first parameter set (black) reproduces the first
insets in both plots, which show the maximum height of peak (I) with measurement position (black in Fig. 2c) well for the first (I) and
rising mobility or carrier concentration, respectively. Even though µ is second (II) peak. Generally, peak (III) seems much broader in the
doubled between each curve in Fig. 2a, the increase in peak height measurement, which might result from frequency-dependent
diminishes with each step, and saturation is reached around 80 cm²/Vs damping γ(ν) for the free charge carriers around the LO-frequency
(yellow curve). In comparison, n2D is only increased linearly and the of the STO substrate41, which is not yet well understood and not
increase in peak height of the relative scattering signal persists across included in the model. The higher background at 900 cm−1 in both
the whole value range shown here. Thus, the characteristic peaks of the measurements compared to the respective simulations could be
near-field fingerprint spectrum are predicted to behave differently for explained by a change in the high-frequency limit ε∞ of the dielectric
the two parameters, finally allowing for direct experimental access to function. Such a change in ε∞ was shown to influence STO near-field
separated n and µ. signals upon doping42,43, possibly also contributing to the broader
Figure 2c presents s-SNOM measurements of the 2DEG finger- peak (III).
print region, for two different positions (black, grey) on LAO/STO For the second parameter set (grey), the goal was to reproduce
with 2DEG (cf. Methods and Supplementary Information for details). the main characteristics of this near-field fingerprint spectrum, i.e., the
For the first position (black curve), a high peak (I) around 750 cm−1, a different intensity distribution between the first two peaks (I) and (II),

Nature Communications | (2025)16:4417 4

Article https://doi.org/10.1038/s41467-025-59633-1

a s-SNOM amplitude b
Position
s2 5
4
3
I) 2
1
II) III)

5 c
4 ×
3 × µ in cm2/Vs
2 ×
1 × 40
× 27
18
I) 12
III)
8
II)
1 µm

Fig. 3 | Fingerprint investigation of lateral inhomogeneities on the nanoscale. spectra shown in Fig. 3b. b Fingerprint spectra measured at positions 1 (blue) to 5
a s-SNOM image of near-field amplitude s2 at ν = 700 cm−1, close to the maximum of (yellow) of Fig. 3a. c Simulation of fingerprint spectra with 2DEG mobility µ rising
the scattering response from the STO phonon near-field resonance (cf. Fig. 1b–d). from 8 cm²/Vs (blue) to 40 cm²/Vs (yellow).
Markers 1–5 indicate measurement positions of a line scan with the resulting

as well as the shift of the minimum between these two peaks to lower Generalized near-field fingerprint spectroscopy
frequencies. However, this matching was not possible by changing the We have shown that near-field fingerprint spectroscopy is a method to
electronic properties alone. Instead, small changes to the LAO/STO investigate the electronic properties of the 2DEG in LAO/STO hetero-
phonon properties were necessary to achieve the frequency shift of structures with high sensitivity, allowing to finally separate influences
the minimum (cf. Supplementary Information S3) and the third max- of n and µ in s-SNOM. Now, we explore the underlying reasons for the
imum (III). This points towards inhomogeneities of the LAO layer, high sensitivity of near-field fingerprint spectroscopy to the 2DEG by
possibly due to differently strained areas after cooling, as strain can evaluating the working principle of s-SNOM step-by-step. In s-SNOM,
lead to a significant shift of the phonon frequencies in oxides44. The the frequency range around zero crossings of Re[ε] has the highest
inhomogeneous 2DEG properties also detected by this method could sensitivity δsn/δε to small changes of the optical properties. Figure 4a
result from magnetic or structural domains observed for LAO/STO shows the calculated s-SNOM coupling function, i.e., the scattering
2DEGs at low temperature45–48. response of the tip-sample coupled system in dependence of the (bulk)
To further investigate how these near-field fingerprint spectra (cf. sample dielectric function ε. Here, the scattering amplitude normal-
Fig. 2c) vary for different positions on the sample, we performed ized to a constant Au-reference s2/s2Au is plotted against Re[ε] in the
preliminary imaging (Fig. 3a) at a frequency of 700 cm−1, where the vicinity of its zero-crossing, with different curves representing differ-
scattering signal is highest due to the strong phonon near-field ent values of the imaginary part, from Im[ε] = 0.5 (light grey) to
response (cf. Fig. 1d, e). These single-frequency s-SNOM images show Im[ε] = 16 (black). The coupling function shows a strongly non-linear
significant variations in amplitude s2 (Fig. 3a) and phase ϕ2 (Supple- behavior between Re[ε] = −10 and +10, with a maximum around −4 and
mentary Information S5). This local inhomogeneity was further inves- a minimum around +1. This behavior is typical for dipolar near-field
tigated with near-field spectra at measurement positions 1–5 indicated coupling and also applicable to the finite dipole model used here. The
by crosses in Fig. 3a. The recorded spectra were again normalized to red-shaded area highlights the range of highest sensitivity δsn/δε, with
the insulating interface and the results are presented in Fig. 3b, a zoom-in shown Fig. 4b. The green arrow indicates a shift to lower
showing the previously discussed fingerprint spectrum of the 2DEG, Re[ε], that would be expected from additional free charge carriers (cf.
with a high (I), a small (II), and a broad peak (III). Interestingly, moving Supplementary Information S3), such as the 2DEG in LAO/STO.
from measurement position 1 (dark blue) to 5 (yellow), the intensity of Beyond LAO/STO, near-field fingerprint spectroscopy is sensitive
the first peak (I) increases by 50%, while the other two peaks remain to different physical effects, such as the LO-frequency of phonons, the
mostly unchanged. This behavior is very similar to Fig. 2a and can be plasma frequency of free charge carriers, or band gaps. This general-
reproduced by a variation of the 2DEG mobility µ between 8 cm²/Vs izability is presented in Fig. 4c–f using the example of doped InAs,
and 40 cm²/Vs (Fig. 3c), showing that with near-field fingerprint which has a doping-dependent zero-crossing of Re[ε] at its “screened”
spectroscopy, the local inhomogeneities observed in the s-SNOM plasma frequency, i.e., shifted compared to the free electron plasma
image (Fig. 3a) can be linked to a laterally varying 2DEG mobility. due to the semiconductor background. Doped InAs also exhibits a
Additional s-SNOM images recorded after the line scan (cf. Supple- surface accumulation layer, which confines a higher concentration of
mentary Information S5) indicate that the inhomogeneous mobility free charge carriers close to the interface to air52. Fig. 4c/d present
could be affected by the previous measurement of the same area. An dielectric functions of a heavily doped (bulk) InAs sample (dark green,
influence of low-temperature scanning probe measurements on the nbulk = 1.6 × 1019 cm−3) as well as its surface accumulation layer (light
electron mobility of LAO/STO is known from literature and a current green, n = 2.0 × 1019 cm−3)53. The higher carrier concentration leads to
topic of research, e.g., arising due to electrostatic gating from the an increase of the screened plasma frequency by Δνp ≈ 22 cm−1, as well
irradiated AFM tip22, persistent photoconductivity49,50, or frozen as lower Re[ε] and higher Im[ε] at any frequency.
condensates51. Using the increased sensitivity of near-field fingerprint The simulated scattering amplitude s2/s2Au of a homogeneously
spectroscopy to the local electronic properties, the physical origin of doped bulk InAs sample is shown in Fig. 4e (dark green). Adding the
such variations can now be investigated in the future. ultrathin (d = 3.6 nm) surface accumulation layer on top of the bulk

Nature Communications | (2025)16:4417 5

Article https://doi.org/10.1038/s41467-025-59633-1

Fig. 4 | Generalization of near-field fingerprint spectroscopy, using the exam- 1019 cm−3). The red shaded area highlights the range of highest s-SNOM sensitivity
ple of surface accumulation layers in doped InAs. a Calculated coupling function δsn/δε. d Zoom-in showing the shift of real and imaginary part. e Near-field ampli-
s2(ε) of s-SNOM for model materials with different imaginary part (light grey to tude spectra s2/s2Au for bulk doped InAs (dark green) and an InAs sample with
black for rising Im[ε]), the red shaded area highlights the range of highest slope. surface accumulation layer of higher carrier concentration (dashed light green).
b Zoom-in highlighting the high sensitivity δsn/δε in this range (cf. Supplementary f Normalized near-field fingerprint spectra showcasing the changes introduced by
Information S2), which is the origin of near-field fingerprint spectra. c Dielectric the surface accumulation layer for different carrier mobilities µ (lighter green). The
function ε(ν) (cf. Supplementary Information S3) of bulk doped InAs (dark peaks (I), (II), and (III) are labeled similarly to the 2DEG in Figs. 1–3.
green, n = 1.6 × 1019 cm−3) and InAs surface accumulation layer (light green, n = 2.0 ×

(dashed light green curve), leads to a small shift of the flank and (minima, maxima, and zero crossings) that improve the quality of
modifications to the shape, similar to the influence of the 2DEG in fitting procedures. For this, it is important to note that a description
LAO/STO (cf. Fig. 1f). Therefore, if a sample with accumulation layer of sample materials and geometry must be built into the theoretical
is normalized to a sample without accumulation layer (Fig. 4f), the model to understand the link between s-SNOM signal and material
near-field fingerprint spectrum shows the characteristic differences properties. As such, additional knowledge from correlative mea-
introduced by the ultrathin surface modification. The InAs finger- surements is necessary to gain meaningful insight into the near-field
print spectrum is very similar to those shown for LAO/STO, with a fingerprint spectrum. The resulting pattern, however, is highly sen-
main peak I) and two smaller peaks II) and III). However, increasing sitive to small changes in optical properties resulting from local
the electron mobility µ (towards lighter green) mainly results in inhomogeneities, surface modifications, or deliberately influencing
changes to peak III) for InAs, whereas it mainly influenced peak I) for the material (e.g., via gating), making s-SNOM with near-field fin-
the LAO/STO 2DEG (cf. Fig. 3), showing that near-field fingerprint gerprint spectroscopy a powerful tool for nanoscale material
spectra are characteristic for the specific material and layer geo- analysis.
metry. This can be used for a generalization to many different
materials, even in bulk materials without surface layers (cf. Supple- Discussion
mentary Information S2). Self-normalization between different states We presented a method to use s-SNOM with novel light sources to
of the material leads to an increased number of characteristic points investigate nanoscale inhomogeneities in materials for modern

Nature Communications | (2025)16:4417 6

Article https://doi.org/10.1038/s41467-025-59633-1

electronic applications, by using near-field fingerprint spectra result- as the first order diffracted spot fulfilled one oscillation, crossing the
ing from self-normalization. These near-field fingerprint spectra are intensity of the specular spot63.
highly sensitive to changes of the electronic properties, due to high
sensitivity δsn/δε to the dielectric function in the fingerprint spectral s-SNOM measurements
range. We used this method to theoretically predict the ideal spectral s-SNOM measurements were done using a commercial low-
region for the s-SNOM investigation of the 2DEG in LAO/STO at low temperature scattering-type scanning near-field optical microscope
temperatures. Three characteristic peaks can be found, which are (cryo-neaSCOPE by attocube systems GmbH), with a pseudo-
differently influenced by 2DEG carrier concentration and mobility. The heterodyne detection configuration64 for single-frequency 2D ima-
changes introduced by 2DEG mobility mostly relate to the main fin- ging and sequential point spectroscopy. As a light source, a commer-
gerprint peak around 750 cm−1, with a saturation of the influence cial narrow-band tunable laser (EKSPLA PT277-XIR) was used,
around µ = 80 cm²/Vs. The influence of carrier density is stronger by consisting of a tunable optical parametric oscillator (OPO) combined
comparison and influences all three peaks similarly. This now allows with difference frequency generation (DFG), that delivers beams of
for the separation of carrier concentration and carrier mobility, which approx. 10 mW power in the DFG region (625–2000 cm−1), in 8 ps
can influence s-SNOM spectra in similar ways. Using a newly-developed pulses with 87 MHz repetition rate and a typical line width of <3 cm−1 in
light source in combination with cryo-s-SNOM at 8 K, we were then this frequency range. As a detector, a photoconductive Mercury Cad-
able to measure near-field fingerprint spectra of the 2DEG by com- mium Telluride (HgCdTe) detector element of 50 µm diameter with a
paring samples of different substrate terminations in a previously ZnSe window was used. As scattering tips, commercial nano-FTIR
inaccessible spectral window with pseudo-heterodyne point spectro- probes by attocube systems GmbH were used, with a tapping ampli-
scopy. We found local variations in measured fingerprint spectra, tude of approximately 80 nm. For in-situ referencing, both samples
which could be explained by changes of the 2DEG and phonon prop- (with and without 2DEG, respectively) were glued side by side on a
erties. We also presented the general applicability of our method by copper plate mounted on the sample stage. The system was cooled
predicting the mobility-dependent fingerprint response of surface down to its base temperature with the samples being stabilized at
accumulation layers in doped InAs. T = 8 K, as monitored by a calibrated Cernox sensor, which was ther-
In the future, this method can be used to disentangle the local mally coupled to the sample plate. Cooling was provided by an inte-
carrier concentration and mobility in investigations of the 2DEG in grated closed-cycle cryostat (attoDry800 from attocube systems
LAO/STO, for example, for the characterization of different types of GmbH), keeping the sample space free of cryogenic media.
defects54, such as varying LAO thickness, atomic steps, or vacancies of
different ions. Additionally, the higher 2DEG mobility at low tem- Theoretical modeling
perature increases the surface sensitivity of s-SNOM compared to The Finite Dipole Model (FDM)35 and Transfer Matrix Method (TMM)
room temperature and should allow for conclusions about the vertical were used in combination36 to determine the near-field scattering
distribution38,55 of the charge carriers. More generally, novel light response of arbitrarily layered systems with known dielectric func-
sources allow fast frequency sweeping across a broad spectral range in tions. FDM parameters were 400 nm ellipsoid length, 90 nm tip radius,
the infrared region with a high signal-to-noise ratio. This enables self- 80 nm tapping amplitude, demodulation order n = 2, and the geo-
normalization even at low scattering amplitudes, as was successfully metric factor g = 0.7 × exp(0.1i). The p-polarized TMM reflection
shown for interband transitions in tetralayer graphene recently56. coefficient was used as the FDM sample reflection factor36, at a domi-
Upon identification of promising characteristic frequencies, s-SNOM nant in-plane wavevector29 of kx = 250 000 cm−1. To model the optical
images can be recorded using a narrow linewidth42, to investigate local properties of the layer stack, the dielectric functions of STO and LAO
defects with nanoscale resolution at the frequency where the contrast were calculated from literature data65,66 using the Berreman-
is strongest. Here, we have demonstrated a link between quantitative Unterwald-Lowndes factorized form41,67 of multiple Lorentz oscilla-
analysis and spatially resolving carrier density and mobility in the 2DEG tors (cf. Supplementary Information S3). To describe the dielectric
of LAO/STO via near-field fingerprint spectra. This method can be function of the 2DEG, a Drude term was added to the STO phonon
easily transferred to other confined correlated electron systems such background41, which was then used in a multilayer approach of 10 sli-
as 2DEGs at different (oxide) heterointerfaces57,58, van-der-Waals het- ces with thickness d = 1 nm and exponentially decaying charge carrier
erostructures or multilayers of 2D materials56,59, or topological concentration away from the interface, distributing the 2D sheet car-
insulators60,61, providing spatially resolved information for designing rier density with a decay constant z0 = 2 nm. The effective mass was
nanoelectronic devices. approximated to an averaged effective mass of m* = 3.2 m0 (with the
free electron mass m0) as predicted by First Principles calculations55
Methods and measured experimentally38.
Sample preparation
Eight unit cells of LaAlO3 were deposited on wet-etched, Data availability
TiO2-terminated (100)-SrTiO3 substrates using pulsed laser deposition Data sets generated during the current study are available from the
(PLD). The films were deposited at a laser frequency of 1 Hz at a fluence corresponding author on request.
of 1.0 J/cm2 and an oxygen pressure of 1 × 10−4 mbar. The growth
temperature was 800 °C and the samples were quenched down to Code availability
room temperature after a relaxation at growth temperature62. During Code used for theoretical calculations in the current study is available
growth, clear RHEED-intensity oscillations were observed, indicating a from the corresponding author on request.
layer-by-layer growth mode and yielding single unit cell thickness
control. The samples were characterized by AC-Hall effect measure- References
ments in van-der-Pauw geometry at room temperature indicating a 1. Hwang, H. Y. et al. Emergent phenomena at oxide interfaces. Nat.
sheet carrier density of 4.3 × 1013 cm−2 and an electron mobility of Mater. 11, 103–113 (2012).
5.1 cm2/Vs. SrO termination was achieved by PLD deposition in the 2. Gunkel, F., Christensen, D. V., Chen, Y. Z. & Pryds, N. Oxygen
same system as for the LAO thin films from a ceramic SrO2 target using vacancies: The (in)visible friend of oxide electronics. Appl. Phys.
a laser fluence of 0.9 J/cm2. The chamber was held at an oxygen partial Lett. 116, 120505 (2020).
pressure 2 × 10−7 mbar and the substrate was heated to 800 °C. To 3. Trier, F. et al. Oxide spin-orbitronics: spin–charge interconversion
achieve the change of termination, the deposition was stopped as soon and topological spin textures. Nat. Rev. Mater. 7, 258–274 (2022).

Nature Communications | (2025)16:4417 7

Article https://doi.org/10.1038/s41467-025-59633-1

4. Schlom, D. G. & Mannhart, J. Interface takes charge over Si. Nat. 31. Taubner, T., Korobkin, D., Urzhumov, Y., Shvets, G. & Hillenbrand, R.
Mater. 10, 168–169 (2011). Near-field microscopy through a SiC superlens. Science 313,
5. Ohtomo, A. & Hwang, H. Y. A high-mobility electron gas at the 1595 (2006).
LaAlO3/SrTiO3 heterointerface. Nature 427, 423–426 (2004). 32. Hillenbrand, R., Taubner, T. & Keilmann, F. Phonon-enhanced light
6. Berner, G. et al. LaAlO3/SrTiO3 oxide heterostructures studied by matter interaction at the nanometre scale. Nature 418, 159–162
resonant inelastic x-ray scattering. Phys. Rev. B 82, 241405 (2010). (2002).
7. Breitschaft, M. et al. Two-dimensional electron liquid state at 33. Mester, L., Govyadinov, A. A., Chen, S., Goikoetxea, M. & Hillen-
LaAlO3-SrTiO3 interfaces. Phys. Rev. B 81, 153414 (2010). brand, R. Subsurface chemical nanoidentification by nano-FTIR
8. Reyren, N. et al. Superconducting interfaces between insulating spectroscopy. Nat. Commun. 11, 3359 (2020).
oxides. Science 317, 1196–1199 (2007). 34. Hensling, F. V. E., Baeumer, C., Rose, M.-A., Gunkel, F. & Dittmann,
9. Li, L., Richter, C., Mannhart, J. & Ashoori, R. C. Coexistence of R. SrTiO3 termination control: a method to tailor the oxygen
magnetic order and two-dimensional superconductivity at LaAlO3/ exchange kinetics. Mater. Res. Lett. 8, 31–40 (2020).
SrTiO3 interfaces. Nat. Phys. 7, 762–766 (2011). 35. Cvitkovic, A., Ocelic, N. & Hillenbrand, R. Analytical model for
10. Richter, C. et al. Interface superconductor with gap behaviour like a quantitative prediction of material contrasts in scattering-type near-
high-temperature superconductor. Nature 502, 528–531 (2013). field optical microscopy. Opt. Express 15, 8550–8565 (2007).
11. Brinkman, A. et al. Magnetic effects at the interface between non- 36. Hauer, B., Engelhardt, A. P. & Taubner, T. Quasi-analytical model for
magnetic oxides. Nat. Mater. 6, 493–496 (2007). scattering infrared near-field microscopy on layered systems. Opt.
12. Ariando, A. et al. Electronic phase separation at the LaAlO3/SrTiO3 Express 20, 13173–13188 (2012).
interface. Nat. Commun. 2, 188 (2011). 37. Son, W., Cho, E., Lee, B., Lee, J. & Han, S. Density and spatial dis-
13. Lee, J.-S. et al. Titanium dxy ferromagnetism at the LaAlO3/SrTiO3 tribution of charge carriers in the intrinsic n-type LaAlO3-SrTiO3
interface. Nat. Mater. 12, 703–706 (2013). interface. Phys. Rev. B 79, 245411 (2009).
14. Keilmann, F. & Hillenbrand, R. Near-field microscopy by elastic light 38. Dubroka, A. et al. Dynamical response and confinement of the
scattering from a tip. Philos. Trans. R. Soc. Lond. A 362, 787–805 electrons at the LaAlO3/SrTiO3 interface. Phys. Rev. Lett. 104,
(2004). 156807 (2010).
15. Chen, X. et al. Modern scattering-type scanning near-field optical 39. Taubner, T., Hillenbrand, R. & Keilmann, F. Nanoscale polymer
microscopy for advanced material research. Adv. Mater. 31, recognition by spectral signature in scattering infrared near-field
e1804774 (2019). microscopy. Appl. Phys. Lett. 85, 5064–5066 (2004).
16. Boltasseva, A. & Shalaev, V. M. Transdimensional photonics. ACS 40. van Mechelen, J. L. M. et al. Electron-phonon interaction and charge
Photonics 6, 1–3 (2019). carrier mass enhancement in SrTiO3. Phys. Rev. Lett. 100,
17. Latil, S. & Henrard, L. Charge carriers in few-layer graphene films. 226403 (2008).
Phys. Rev. Lett. 97, 36803 (2006). 41. Gervais, F., Servoin, J.-L., Baratoff, A., Bednorz, J. G. & Binnig, G.
18. Gunkel, F. et al. Space charges and defect concentration profiles at Temperature dependence of plasmons in Nb-doped SrTiO3. Phys.
complex oxide interfaces. Phys. Rev. B 93, 245431 (2016). Rev. B 47, 8187–8194 (1993).
19. Rose, M.-A. et al. Identifying ionic and electronic charge transfer at 42. Lewin, M. et al. Nanospectroscopy of infrared phonon resonance
oxide heterointerfaces. Adv. Mater. 33, 2004132 (2021). enables local quantification of electronic properties in doped
20. Park, D.-S. et al. The emergence of magnetic ordering at complex SrTiO3 ceramics. Adv. Funct. Mater. 28, 1802834 (2018).
oxide interfaces tuned by defects. Nat. Commun. 11, 3650 (2020). 43. Qin, T.-X. et al. Revealing the interaction of charge carrier-phonon
21. Cheng, L. et al. Near-field imaging of the LaAlO3/SrTiO3 interfacial coupling by quantification of electronic properties at the SrTiO3/
conductivity. J. Infrared Millim. Waves 36, 542–546 (2017). TiO2 heterointerface. Nano Lett. 22, 2755–2761 (2022).
22. Luo, W. et al. High sensitivity variable-temperature infrared nano- 44. Kostov, K. L., Schumann, F. O., Polzin, S., Sander, D. & Widdra, W.
scopy of conducting oxide interfaces. Nat. Commun. 10, 2774 NiO growth on Ag(001): A layer-by-layer vibrational study. Phys.
(2019). Rev. B 94, 075438 (2016).
23. Barnett, J. et al. Phonon‐enhanced near‐field spectroscopy to 45. Christensen, D. V. et al. Strain-tunable magnetism at oxide domain
extract the local electronic properties of buried 2D electron sys- walls. Nat. Phys. 15, 269–274 (2019).
tems in oxide heterostructures. Adv. Funct. Mater. 30, 2004767 46. Honig, M. et al. Local electrostatic imaging of striped domain order
(2020). in LaAlO3/SrTiO3. Nat. Mater. 12, 1112–1118 (2013).
24. Rose, M.-A. et al. Local inhomogeneities resolved by scanning 47. Noad, H., Wittlich, P., Mannhart, J. & Moler, K. A. Modulation of
probe techniques and their impact on local 2DEG formation in oxide superconducting transition temperature in LaAlO3/SrTiO3 by
heterostructures. Nanoscale Adv. 3, 4145–4155 (2021). SrTiO3 structural domains. J. Supercond. Nov. Magn. 32, 821–825
25. Taubner, T., Keilmann, F. & Hillenbrand, R. Nanoscale-resolved (2019).
subsurface imaging by scattering-type near-field optical micro- 48. Kalisky, B. et al. Locally enhanced conductivity due to the tetra-
scopy. Opt. Express 13, 8893–8899 (2005). gonal domain structure in LaAlO3/SrTiO3 heterointerfaces. Nat.
26. Krutokhvostov, R. et al. Enhanced resolution in subsurface near- Mater. 12, 1091–1095 (2013).
field optical microscopy. Opt. Express 20, 593–600 (2012). 49. Irvin, P. et al. Rewritable nanoscale oxide photodetector. Nat. Pho-
27. Govyadinov, A. A. et al. Recovery of permittivity and depth from tonics 4, 849–852 (2010).
near-field data as a step toward infrared nanotomography. ACS 50. Eom, K. et al. Origin of the giant persistent photoconductivity in
Nano 8, 6911–6921 (2014). LaAlO3/SrTiO3 heterostructures probed by noise spectroscopy. J.
28. Jung, L. et al. Quantification of carrier density gradients along axi- Mater. Sci. Technol. 137, 152–158 (2023).
ally doped silicon nanowires using Infrared Nanoscopy. ACS Pho- 51. Bi, F. et al. Water-cycle” mechanism for writing and erasing
tonics 6, 1744–1754 (2019). nanostructures at the LaAlO3/SrTiO3 interface. Appl. Phys. Lett. 97,
29. Fei, Z. et al. Infrared nanoscopy of dirac plasmons at the graphene- 173110 (2010).
SiO2 interface. Nano Lett. 11, 4701–4705 (2011). 52. Bell, G. R., Jones, T. S. & McConville, C. F. Accumulation layer
30. Dai, S. et al. Tunable phonon polaritons in atomically thin van der profiles at InAs polar surfaces. Appl. Phys. Lett. 71, 3688–3690
Waals crystals of boron nitride. Science 343, 1125–1129 (2014). (1997).

Nature Communications | (2025)16:4417 8

Article https://doi.org/10.1038/s41467-025-59633-1

53. Jung, L. Mid-infrared near-field investigation of resonances in Author contributions

doped semiconductors (PhD thesis, RWTH Aachen University, J.B., T.T., and F.G. conceived and planned the study. J.B., Y.C.D., and R.H.
Aachen, 2020). planned and prepared the measurements, which J.B., T.T., Y.C.D., and
54. Li, W., Shi, J., Zhang, K. H. L. & MacManus-Driscoll, J. L. Defects in R.H. carried out. M.A.R. and F.G. provided the samples; J.B., M.A.R., and
complex oxide thin films for electronics and energy applications: F.G. did pre- and post-characterization. J.B. and K.W. performed data
challenges and opportunities. Mater. Horiz. 7, 2832–2859 (2020). evaluation and simulations. All authors discussed the data and con-
55. Delugas, P. et al. Spontaneous 2-dimensional carrier confinement at tributed to writing the paper. All authors approved the final version of
the n-type SrTiO3/LaAlO3 interface. Phys. Rev. Lett. 106, the manuscript.
166807 (2011).
56. Wirth, K. G. et al. Experimental observation of ABCB stacked tetra- Funding
layer graphene. ACS Nano 16, 16617–16623 (2022). Open Access funding enabled and organized by Projekt DEAL.
57. Mannhart, J., Blank, D., Hwang, H. Y., Millis, A. J. & Triscone, J.-M.
Two-dimensional electron gases at oxide interfaces. MRS Bull. 33, Competing interests
1027–1034 (2008). Y.C.D. and R.H. are employees of attocube systems GmbH, producer of
58. Stemmer, S. & James Allen, S. Two-dimensional electron gases at the cryo-s-SNOM microscope used in this study. The remaining authors
complex oxide interfaces. Annu. Rev. Mater. Res. 44, 151–171 (2014). declare no competing interests.
59. Castellanos-Gomez, A. et al. Van der Waals heterostructures. Nat.
Rev. Methods Prim. 2, 58 (2022). Additional information
60. Pogna, E. A. A. et al. Mapping propagation of collective modes in Supplementary information The online version contains
Bi2Se3 and Bi2Te2.2Se0.8 topological insulators by near-field ter- supplementary material available at
ahertz nanoscopy. Nat. Commun. 12, 6672 (2021). https://doi.org/10.1038/s41467-025-59633-1.
61. Wu, J.-S., Basov, D. N. & Fogler, M. M. Topological insulators are
tunable waveguides for hyperbolic polaritons. Phys. Rev. B 92, Correspondence and requests for materials should be addressed to
205430 (2015). Thomas Taubner.
62. Xu, C. et al. Disentanglement of growth dynamic and thermo-
dynamic effects in LaAlO3/SrTiO3 heterostructures. Sci. Rep. 6, Peer review information Nature Communications thanks the anon-
22410 (2016). ymous reviewer(s) for their contribution to the peer review of this work. A
63. Sun, H. Y. et al. Chemically specific termination control of oxide peer review file is available.
interfaces via layer-by-layer mean inner potential engineering. Nat.
Commun. 9, 2965 (2018). Reprints and permissions information is available at
64. Ocelic, N., Huber, A. & Hillenbrand, R. Pseudoheterodyne detection http://www.nature.com/reprints
for background-free near-field spectroscopy. Appl. Phys. Lett. 89,
101124 (2006). Publisher’s note Springer Nature remains neutral with regard to jur-
65. Kamarás, K. et al. The low‐temperature infrared optical functions of isdictional claims in published maps and institutional affiliations.
SrTiO3 determined by reflectance spectroscopy and spectroscopic
ellipsometry. J. Appl. Phys. 78, 1235–1240 (1995). Open Access This article is licensed under a Creative Commons
66. Willett-Gies, T., DeLong, E. & Zollner, S. Vibrational properties of Attribution 4.0 International License, which permits use, sharing,
bulk LaAlO3 from Fourier-transform infrared ellipsometry. Thin adaptation, distribution and reproduction in any medium or format, as
Solid Films 571, 620–624 (2014). long as you give appropriate credit to the original author(s) and the
67. Schöche, S. et al. Infrared dielectric functions, phonon modes, and source, provide a link to the Creative Commons licence, and indicate if
free-charge carrier properties of high-Al-content AlxGa1−xN alloys changes were made. The images or other third party material in this
determined by mid infrared spectroscopic ellipsometry and optical article are included in the article’s Creative Commons licence, unless
Hall effect. J. Appl. Phys. 121, 205701 (2017). indicated otherwise in a credit line to the material. If material is not
included in the article’s Creative Commons licence and your intended
Acknowledgements use is not permitted by statutory regulation or exceeds the permitted
J.B., K.W., and T.T. acknowledge funding from the Deutsche For- use, you will need to obtain permission directly from the copyright
schungsgesellschaft (DFG) within the collaborative research center SFB holder. To view a copy of this licence, visit http://creativecommons.org/
917 and under grant agreement no. TA 848/7-1. M.A.R. and F.G. licenses/by/4.0/.
acknowledge financial support from the Deutsche For-
schungsgesellschaft (DFG) FG 1604 (No. 315025796). © The Author(s) 2025

Nature Communications | (2025)16:4417 9