Miniaturized Spectrometers with a Tunable van der
Waals Junction
Hoon Hahn Yoon,1,2,∗ Henry A. Fernandez,1,2 Fedor Nigmatulin,1,2
Weiwei Cai,3 Zongyin Yang,4 Hanxiao Cui,5 Faisal Ahmed,1 Xiaoqi Cui,1,2
Md Gius Uddin,1,2 Ethan D. Minot,6 Harri Lipsanen,1 Kwanpyo Kim,7
Pertti Hakonen,2 Tawfique Hasan,8 Zhipei Sun1,2,∗
1Department of Electronics and Nanoengineering, Aalto University, Espoo 02150, Finland
2QTF Centre of Excellence, Department of Applied Physics, Aalto University,
Aalto 00076, Finland
3Key Lab of Education Ministry for Power Machinery and Engineering,
School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China
4College of Information Science and Electronic Engineering,
State Key Laboratory of Modern Optical Instrumentation,
Zhejiang University, Hangzhou 310027, China
5School of Aeronautics and Astronautics, Sichuan University, Chengdu 610065, China
6Department of Physics, Oregon State University, Corvallis, Oregon 97331, United States
7Department of Physics, Yonsei University, Seoul 03722, Republic of Korea
8Cambridge Graphene Centre, University of Cambridge, Cambridge CB3 0FA, UK
∗To whom correspondence should be addressed;
E-mail: hoonhahn.yoon@aalto.fi and zhipei.sun@aalto.fi
Miniaturized computational spectrometers, which can obtain incident spectra
using a combination of device spectral response and reconstruction algorithm,
are essential for on-chip and implantable applications. Highly-sensitive spec-
tral measurement using a single detector allows the footprints of such spec-
trometers to be scaled down while achieving spectral resolution approach-
1
ing that of benchtop systems. We report a high-performance computational
spectrometer based on a single van der Waals junction with an electrically-
tunable transport-mediated spectral response. We achieve high peak wave-
length accuracy (∼ 0.36 nm), high spectral resolution (∼ 3 nm), broad opera-
tion bandwidth (from ∼ 405 to 845 nm), and proof-of-concept spectral imag-
ing. Our approach provides a route towards ultra-miniaturization and offers
unprecedented performance in accuracy, resolution, and operation bandwidth
for single-detector computational spectrometers.
Spectrometers are indispensable for various applications, including industrial inspection,
chemical/biological characterization, and image sensing/analysis (1, 2). Their miniaturization
with high spectral resolution and wide operation bandwidth is highly desirable to meet the
emerging and future demands in portable and on-chip applications (1). However, conventional
spectroscopy systems typically rely on bulky dispersive optical components (e.g., gratings) and
detector or filter arrays, which impose strict restrictions on ultra-miniaturization (2).
Common spectrometer miniaturization approaches therefore replace the functions of these
dispersive optical elements through various schemes (Fig. S1), including photonic crystals (3),
metasurfaces (4), and compact interferometers (5). Recently, a profound technological leap has
seen the emergence of miniaturized computational spectrometers, which leverage the power of
mathematical algorithms for spectrum reconstruction (1). Examples of such approaches have
used quantum dot filter arrays on to the charged-coupled device sensor (6), bandgap engineered
multiple nanowires (7), a single nanowire with bandgap gradation (8), Stark effect in black
phosphorus (9), in situ perovskite modulation (10), and a single superconducting nanowire with
tunable quantum efficiency (11). However, the performance and usability of these computa-
tional spectrometers remain limited: spectral resolution and operation bandwidth are typically
restricted by the number of integrated detectors (6–8), bandgap modulation limits (9, 10), and
2
cryogenic operational requirements (11).
Photodetection with two-dimensional (2D) layered materials is advantageous due to their
strong light-matter interaction, atomically-sharp interface, and electrically-tunable photoresponse
(12–14). However, insufficient band structure modulation makes it challenging to achieve high-
resolution, broadband spectral sensing using a single 2D material. On the other hand, 2D
materials-based van der Waals (vdW) junctions offer highly tunable functionalities beyond the
constituent materials (15–17) and could overcome these limitations. In particular, we suggest
that wavelength-dependent photodetection with vdW junctions recently exploited for optoelec-
tronic logic computing (18, 19) and color sensing (20, 21) could also be key to high-resolution
computational spectral sensing.
Here, we demonstrate a high-performance ultra-miniaturized computational spectrometer
utilizing a single vdW junction with an electrically-tunable transport-mediated spectral re-
sponse. Our device, with its footprint defined by the junction size (∼ 22 × 8 µm2) shows
unprecedented performance for a single detector computational spectrometer, with the ability
to resolve peak monochromatic wavelengths with ∼ 0.36 nm accuracy, reconstruct broadband
spectra with ∼ 3 nm resolution, and acquire spectral images by scanning. Our single-junction
spectrometer concept can be extended to other tunable junctions for achieving high spectral
resolution and broad operation bandwidth with its ultra-compact size, representing the ultimate
miniaturization strategy without sacrificing spectrometer performance.
The performance of computational spectrometers relies on the variability of their wavelength-
dependent photoresponsivity (1,6–11). The single-detector miniaturized spectrometers reported
thus far are limited by their performance (9, 10) and usability (9, 11) due to the limited band
structure modulation and consequently, the spectral response. In contrast, electrical tuning of
the interfacial band alignment of a vdW junction (Fig. 1A, upper panel) enables controllable
and unique interlayer transport (15–17). This allows a tunable spectral response (Fig. 1A, lower
3
panel) with high sensitivity and variability over a wide spectral range (12–14). This suggests
that a single-vdW-junction spectrometer could achieve significantly higher performance than
previously reported spectrometers (section ST1 and table S1). We combine an electrically-
tunable single-vdW-junction with computational reconstruction algorithms for various applica-
tions (Fig. 1B). To experimentally realize our spectrometer concept, the following three steps
are required (Fig. S2): measuring the gate-tunable spectral responses with multiple known in-
cident spectra (learning process); measuring the gate-tunable photocurrent of the unknown in-
cident light to be analyzed (testing process); and computing the spectral information of the
Spectral imager
A
Electrically-tunable spectral response matrix
Electrically-tunable vdW junction B Wavelength meter
In
te
n
s
it
y
Wavelength
Monochromatic light
Gate tuning
R
e
s
p
o
n
s
iv
it
y
W
avelength
E
v
E
c
Spectrometer
In
te
n
s
it
y
Wavelength
Broadband light
W
a
v
e
le
n
g
th
Image
Figure 1: Ultra-miniaturized spectrometer concept with a single-vdW-junction. A, A typ-
ical gate-tunable band alignment at the vdW junction interface (upper) with its distinct gate-
tunable spectral response matrix (lower). EC (EV) represents the conduction (valence) band
edge. B, Schematic of various application examples using single-junction spectrometer: wave-
length meter to distinguish peak wavelengths of monochromatic light (upper), spectrometer
to resolve broadband spectra (middle), and spectral imager to analyze spectral information of
images (lower).
4
unknown incident light based on the results obtained in learning and testing processes with the
reconstruction algorithm (reconstructing process).
The distinct and varied photoresponse of a vdW junction, tuned at different gate voltages
and incident light wavelengths, is critical to our spectrometer (1). We choose a MoS2/WSe2
heterojunction (Fig. 2A) as an example for its distinct spectral response due to the gate-tunable
photovoltaic effect from the visible to the near-infrared. (22–28) The MoS2/WSe2 heterojunc-
tion is encapsulated by top and bottom h-BN layers for insulation and passivation, respectively
(section MM1). A monolayer graphene film below the stack is used as a local gate electrode
for effective gate tuning. Each stacking layer was characterized by Raman spectroscopy and
atomic force microscopy (Fig. S3) to confirm the quality of the vdW heterostructure.
The transfer curves (drain-source current IDS as a function of the gate-source voltage V GS)
of the MoS2 or WSe2 channels and their heterojunction are measured at drain-source voltage
V DS = 3 V in dark condition (Fig. 2B). The individual MoS2 and WSe2 channels exhibit n(p)-
type characteristics due to the donor (acceptor) impurities in MoS2 (WSe2). Thus, a depletion
region and built-in electric field are expected at their vdW interface (22–28). The MoS2/WSe2
heterojunction is characterized by positive V DS applied to the WSe2 side, corresponding to
the forward biasing of the diode. The sign change of transconductance, dIDS
dVGS
, occurs at V GS
= ∼ -5 V, matching the hole current from WSe2 with the electron current from MoS2. This
“anti-ambipolar” behavior and other transport properties (Figs. S4 to S7) are typical of the
MoS2/WSe2 heterojunctions (22–28), providing clearly distinguishable V GS-dependence.
The transfer curves of the MoS2/WSe2 heterojunction measured under multiple known inci-
dent lights with a bandwidth of∼ 10 nm indicate a strong wavelength dependence (Fig. 2C). The
photoresponsivity, R =
Iph
P
, measured at different V GS and incident light wavelengths is used
to encode the spectral response matrix, where the photocurrent is defined as Iph = I light − Idark,
with I light and Idark representing IDS with and without light illumination at V DS = 3 V, respec-
5
WSe
2V
DS
V
GS
h-BN
MoS
2
Graphene
50E G
A
C D
F
Quasi-monochromatic spectra (E)
Complex broadband spectra (F)
100 50 25 12.5
0
10
20
30
40
P
e
a
k
 s
ig
n
a
l-
to
-n
o
is
e
 r
a
ti
o
 (
d
B
)
Learning step (nm)
λ (nm)
   845
   800
   750
   700
   650
   600
   550
   500
   450
   405
I D
S
 (
A
)
Wavelength (nm)
-13 R (mA·W-1) 0
-20
-10
0
10
20
400 500 600 700 800
V
G
S
 (
V
)
-20 -10 0 10
10
-9
10
-8
10
-7
V
DS
 = 3 V
MoS
2
/WSe
2
V
GS
 (V)
B
-30 -20 -10 0 10 20 30
10
10
10
10
MoS
2
MoS
2
/WSe
2
WSe
2
I D
S
 (
A
)
V
DS
 = 3 V
V
GS
 (V)
0.0
0.4
P
 (
a
.u
.)
20
400 500 600 700 800
0.8
0.0
0.4
1.2
P
 (
a
.u
.)
Commercial spectrometer
Single-junction spectrometer
Wavelength (nm) Wavelength (nm)
400 500 600 700 800
0.8
0.0
0.4
0.8
1.2
10 μm
Top
h-BN
MoS
2
WSe
2
Graphene
10 μm
Bottom
h-BN
-10
-8
-6
-4
Figure 2: Single-junction spectrometer demonstration. A, Schematic of our MoS2/WSe2
heterojunction spectrometer (left) and its optical images on the h-BN and graphene layers before
(middle) and after (right) depositing electrodes and stacking the top h-BN passivation layer. The
top h-BN layer is not present in the upper panel for better visibility. B and C, Transfer curves of
the MoS2 and WSe2 channels and their heterojunction with the graphene gate without (B) and
with (C) light illumination at different wavelengths with a fixed power of ∼ 20 µW. D, Color
contour plot of the spectral response matrix. E and F, Quasi-monochromatic (E, bandwidth:
∼ 10 nm), two different broadband (F) spectra reconstructed with our spectrometer (solid) and
measured using a commercial spectrometer (dashed). G, Peak signal-to-noise ratio between
reconstructed and reference spectra as a function of learning step.
6
tively, and P is the incident light power (Fig. S8). The gate-tunable spectral response with
high sensitivity over a wide spectral range is due to the wavelength-dependent absorption (29)
of MoS2 and WSe2 as well as the controllable charge carrier transport (22–28) through the
MoS2/WSe2 interface, unlike the individual material. The spectral response matrix (Fig. 2D)
inherits a rich structure from the dynamics of photoexcited charge carriers generated across the
tunable MoS2/WSe2 heterojunction (22–29), confirming fast and stable spectral detection with
giant gate-tunability in our MoS2/WSe2 heterojunction (Figs. S9 to S13).
After encoding this spectral response matrix (Fig. 2D) for the learning process, our single-
junction spectrometer is ready to measure unknown incident light spectra, following the work-
flow diagram (Fig. S2). Briefly, we measure the gate-tunable photocurrent of the unknown in-
cident light and then compute its constrained least-square solution to reconstruct the spectrum
using adaptive Tikhonov regularization by minimizing the residual norm with a regularization
factor (1, 8). Details of the optical setup, electrical/optoelectrical measurements, and computa-
tional reconstruction are provided in Fig. S14 and sections MM2 to MM4.
The quasi-monochromatic and complex broadband spectra reconstructed with our single-
junction spectrometer agree well with the reference spectra measured using a commercial spec-
trometer, demonstrating the viability of single-junction spectrometer concept (Fig. 2, E and F).
While the demonstrated bandwidth (from ∼ 405 to 845 nm) is limited due to the availability
of the light wavelengths in our laboratory, the MoS2/WSe2 heterojunction exhibits photore-
sponse from ∼ 400 to 2400 nm (25). Indeed, the vdW junctions are known to exhibit pho-
todetection capability for incident light whose wavelength corresponds to around half (or even
much smaller than) the bandgap of each material (15–17, 25). Therefore, in principle, single-
junction spectrometer is not limited by the material bandgap and likely offers operation band-
width broader than our demonstration. Our single-junction spectrometer using the interlayer-
transport-mediated photoresponse is fundamentally different from the previously demonstrated
7
spectrometer concepts, such as bandgap engineering and grading (7–10). Detailed compar-
isons of our work with the current state-of-the-art miniaturized spectrometers (including black-
phosphorus-based spectrometers with the Stark effect (9)) are given in section ST2 and table
S1.
To evaluate deviations between the reconstructed and reference spectra, the peak signal-to-
noise ratio (PSNR) has previously been used to analyze the mean squared error (section ST3).
The maximum PSNR estimated from the extrapolation is ∼ 35.7 and 33.6 dB for the quasi-
monochromatic and complex broadband spectra, respectively (Fig. 2G). A reasonable learning
step (i.e. step in wavelength for encoding the spectral response matrix) can be chosen based on
the saturated PSNR. Therefore, a high-speed learning process is achievable with a large learning
step and slightly reduced accuracy (11).
The wavelength resolving power is an important measure of spectrometers in practical ap-
plications (1, 2). To demonstrate high spectral resolution capability with our single-junction
ultra-miniaturized spectrometer, we construct a high-density spectral response matrix through
an ultra-small learning step of ∼ 0.1 nm using monochromatic light from ∼ 675 nm to 685
nm for the learning process (Fig. 3A). Our single-junction spectrometer encoded by the high-
density spectral response matrix can resolve monochromatic light with high accuracy (Fig. 3,
B and C). The average peak wavelength difference (∆λ) between reconstructed and reference
spectra is ∼ 0.36 ± 0.06 nm, with a minimum of ∼ 0.04 nm (Fig. 3D). This is comparable to
the learning step of ∼ 0.1 nm. The wavelength resolving power
(
Rλ =
λ
∆λ
)
at a given input
wavelength (λ) is ∼ 3470 ± 880.
Furthermore, we measure complex incident spectra to study spectral resolution. Two peaks
at ∼ 679 nm, separated by ∼ 3 nm, are successfully distinguished (Fig. 3E). Our spectrometer
can also resolve broadband spectra and identify their peak wavelengths with high resolution
(∼ 0.9 nm demonstrated in Fig. S15). This indicates that our spectrometer has spectral resolu-
8
tion comparable to or better than the current state-of-the-art miniaturized spectrometers (1–11)
with footprint (∼ 22 × 8 µm2) comparable to or smaller than most. This is several orders of
magnitude smaller than commercial miniaturized spectrometers (30) and recently demonstrated
spectrometers with metasurfaces (4), quantum dots (6), or a single-dot perovskite (10); see Ta-
ble S1 for a detailed comparison. We note that the demonstrated accuracy (∼ 0.36 nm) and
103
16
C
676 678 680 682 684
676
678
680
682
684
P
e
a
k
 w
a
v
e
le
n
g
th
(n
m
)
Input wavelength (nm)
A B
676 678 680 682 684676 678 680 682 684
-20
-10
0
10
20
-17 R (mA·W-1) 0
V
G
S
 (
V
)
Wavelength (nm)
676 678 680 682 684
3.0 nm
D F
NPDR change per wavelength (W-1·nm-1)
Commercial
spectrometer
 R = 1 A·W-1
0.8
0.0
0.4
1.2
P
(a
.u
.)
0.8
0.0
0.4
1.2
P
(a
.u
.)
E
0.0
0.4
0.8
1.2
Δ
λ
(n
m
)
676 678 680 682 684
102
103
104
R
e
s
o
lv
in
g
p
o
w
e
r
Input wavelength (nm)
Commercial spectrometer
Single-junction spectrometer
Commercial spectrometer
Single-junction spectrometer
Commercial spectrometer
Single-junction spectrometer
Wavelength (nm)
Wavelength (nm)
S
p
e
c
tr
a
l
re
s
o
lu
ti
o
n
(n
m
)
1013 1014 1015 10
10-5
10-3
10-1
101
 R = 0.1 A·W-1
Achieved in this work
Achievable
with improvement
Figure 3: High-performance wavelength resolving power and spectral resolution. A, Color
contour plot of the high-density spectral response matrix with a learning step of ∼ 0.1 nm. B,
Monochromatic (bandwidth: ∼ 2 nm) spectra reconstructed with our spectrometer (solid) and
measured using a commercial spectrometer (dashed). C, Peak wavelengths of the reconstructed
and measured spectra as a function of input wavelength. D, Peak wavelength difference be-
tween reconstructed and reference spectra (upper panel), and wavelength resolving power of
our single-junction spectrometer (lower panel). E, Complex spectra reconstructed (solid) and
measured (dashed). F, Future prospect of our single-junction spectrometer aiming for ultra-
high-resolution.
9
resolution (∼ 3 nm), which are limited by the smallest incident wavelength step available in our
laboratory, can be further improved by minimizing the learning step during the learning process.
We suggest that such a learning process is practical for applications, similar to the calibration
process in traditional spectrometers.
Many strategies can be considered to improve the resolution, accuracy, and speed of our
single-junction spectrometer (1). These include: (A) increasing the dataset size to create higher-
density spectral response matrix by minimizing the learning step (7–11), (B) designing junctions
with higher response and larger wavelength or gate dependence (15–17), and (C) optimizing the
reconstruction algorithm (e.g., suppressing the perturbation with more advanced regularization
(1, 8) or increasing the accuracy with convolutional processing (20, 21)). Ideally, decreasing
the learning step is a straightforward approach to forming a denser spectral response matrix
for more accurate spectral reconstruction. However, there is a trade-off: signal difference,
measured at small learning steps, comparable to the measurement noise could result in errors
during reconstruction.
To illustrate the future development possibilities of single-junction spectrometers, we con-
sider the normalized photocurrent-to-dark current ratio (NPDR) change per wavelength step
of two resolved peaks (section ST4). The extrapolated line in Fig. 3F indicates the potential
of our approach with improved photoresponsivity for higher resolution than the commercial
miniaturized spectrometers (30). The achievable resolution is highlighted based on recently
reported photoresponse (∼ 0.1 to 1 A·W−1 at 532 nm) of MoS2/WSe2 heterojunctions (25).
The resolution and bandwidth can be further improved by engineering junctions with different
combinations of various 2D materials or integrating waveguides (15–17). Additional strategies
for improving performance are provided in section ST5. With a significant potential to improve
performance, our single-junction spectrometer can not only be adapted to other tunable junc-
tion architectures but also integrated with CMOS(complementary metal-oxide-semiconductor)-
10
compatible platforms.
Our single-junction spectrometer can benefit from the recently developed large-scale 2D
material synthesis to construct an array for future spectral imaging. Here, we demonstrate proof-
of-concept spectral imaging of a color filter consisting of red, blue, and transparent areas with
spatial scanning using our spectrometer (Fig. 4A). At each mapping position, the measured pho-
tocurrent data at different V GS are recorded in the spatial response data cube for spectral recon-
struction. A series of photocurrent mapping data scanned at different V GS is displayed (Fig. 4B)
and converted to a series of spectral data reconstructed at different wavelengths (Fig. 4C). The
B
C
-10 V 10 V7.5 V5 V2.5 V-2.5 V-5 V-7.5 V
I p
h
 (
a
.u
.)
405 nm 580 nm450 nm 500 nm 550 nm 800 nm630 nm 650 nm
P
 (
a
.u
.)
Photocurrent data cube
I
ph 
(x, y, V
GS
)
x
y
V
GS
Spectral data cube
P (x, y, λ)
x
y
λ
Spectral
reconstruction
Image scanning
Single-junction
spectrometer
x
y
Broadband
light source
Transmission
mapping
Near-infraredVisible
A
Figure 4: Proof-of-concept demonstration of spectral imaging. A, Configuration of spectral
imaging using our single-junction spectrometer with a spatial scanning method. A broadband
light source filtered with a color image is incident to our spectrometer for spectral imaging.
B, Photocurrent mapping data scanned at different V GS. C, Spectral images reconstructed at
different wavelengths, covering the visible to near-infrared ranges. Higher intensity at each
wavelength indicates more broadband light is transmitted through the color image. The pixel
intensity in B and C is normalized with each maximum intensity.
11
spectral images indicate that the red and blue filters absorb more incident broadband light from
∼ 405 to 580 nm, and from ∼ 600 to 700 nm, respectively. As a result, the spectra of the red
uppercase alphabet "A" (from∼ 450 to 700 nm) and the blue exclamation mark "!" (from∼ 405
to 845 nm) are distinguishable from that of the background. Note that a strong light signal at
near 800 nm for the exclamation mark "!" can be fully detected, highlighting the advantages of
spectral imaging over conventional RGB color imaging (Fig. S16 with different color filters). In
our demonstration, the image resolution is defined by the mapping step. However, our concept
has great potential for large-scale spectral imaging by future array devices, offering high spatial
resolution with the junction at the micrometer- or nanometer-scale.
In our spectrometer, no photodetector array, filter array, or other bulky dispersive compo-
nents are required to achieve high resolution, sub-nm accuracy, and broad operation bandwidth.
The compact footprint of our single-junction spectrometers may provide scalability and com-
patibility with the current photonic integrated circuits and CMOS-compatible processes for di-
rect integration into modern smartphones, lab-on-a-chip systems, and other customized devices
ranging from bio-implants to drones and satellites.
References
1. Z. Yang, T. Albrow-Owen, W. Cai, T. Hasan, Miniaturization of optical spectrometers.
Science 371, eabe0722 (2021).
2. R. F. Wolffenbuttel, State-of-the-art in integrated optical microspectrometers. IEEE Trans.
Instrum. Meas. 53, 197-202 (2004).
3. Z. Wang et al., Single-shot on-chip spectral sensors based on photonic crystal slabs. Nat.
Commun. 10, 1020 (2019).
12
4. A. Tittl et al., Imaging-based molecular barcoding with pixelated dielectric metasurfaces.
Science 360, 1105-1109 (2018).
5. D. M. Kita et al., High-performance and scalable on-chip digital Fourier transform spec-
troscopy. Nat. Commun. 9, 4405 (2018).
6. J. Bao, M. G. Bawendi, A colloidal quantum dot spectrometer. Nature 523, 67-70 (2015).
7. J. Meng, J. J. Cadusch, K. B. Crozier, Detector-only spectrometer based on structurally
colored silicon nanowires and a reconstruction algorithm. Nano Lett. 20, 320-328 (2019).
8. Z. Yang et al., Single-nanowire spectrometers. Science 365, 1017-1020 (2019).
9. S. Yuan, D. Naveh, K. Watanabe, T. Taniguchi, F. Xia, A wavelength-scale black phospho-
rus spectrometer. Nat. Photonics 15, 601-607 (2021).
10. L. Guo et al., A single-dot perovskite spectrometer. Adv. Mater., 2200221 (2022).
11. L. Kong et al., Single-detector spectrometer using a superconducting nanowire. Nano Lett.
21, 9625-9632 (2021).
12. F. Xia, T. Mueller, Y.-m. Lin, A. Valdes-Garcia, P. Avouris, Ultrafast graphene photodetec-
tor. Nat. Nanotechnol. 4, 839-843 (2009).
13. M. Buscema et al., Photocurrent generation with two-dimensional van der Waals semicon-
ductors. Chem. Soc. Rev. 44, 3691-3718 (2015).
14. F. Koppens et al., Photodetectors based on graphene, other two-dimensional materials and
hybrid systems. Nat. Nanotechnol. 9, 780-793 (2014).
15. A. K. Geim, I. V. Grigorieva, Van der Waals heterostructures. Nature 499, 419-425 (2013).
13
16. M. M. Furchi, A. Pospischil, F. Libisch, J. Burgdörfer, T. Mueller, Photovoltaic effect in an
electrically tunable van der Waals heterojunction. Nano Lett. 14, 4785-4791 (2014).
17. N. Ubrig et al., Design of van der Waals interfaces for broad-spectrum optoelectronics. Nat.
Mater. 19, 299-304 (2020).
18. S. Wang et al., Nonvolatile van der Waals heterostructure phototransistor for encrypted
optoelectronic logic circuit. ACS Nano 16, 4528-4535 (2022).
19. Y. Sun, Y. Ding, D. Xie, Mixed-dimensional van der Waals heterostructures enabled opto-
electronic synaptic devices for neuromorphic applications. Adv. Funct. Mater. 31, 2105625
(2021).
20. S. Lee, R. Peng, C. Wu, M. Li, Programmable black phosphorus image sensor for broad-
band optoelectronic edge computing. Nat. Commun. 13, 1485 (2022).
21. L. Pi et al., Broadband convolutional processing using band-alignment-tunable heterostruc-
tures. Nat. Electron. 5, 248-254 (2022).
22. C.-H. Lee et al., Atomically thin p–n junctions with van der Waals heterointerfaces. Nat.
Nanotechnol. 9, 676-681 (2014).
23. T. Roy et al., Dual-gated MoS2/WSe2 van der Waals tunnel diodes and transistors. ACS
Nano 9, 2071-2079 (2015).
24. A. Nourbakhsh, A. Zubair, M. S. Dresselhaus, T. Palacios, Transport properties of a
MoS2/WSe2 heterojunction transistor and its potential for application. Nano Lett. 16, 1359-
1366 (2016).
25. M. Long et al., Broadband photovoltaic detectors based on an atomically thin heterostruc-
ture. Nano Lett. 16, 2254-2259 (2016).
14
26. Y. Son et al., Observation of switchable photoresponse of a monolayer WSe2–MoS2 lateral
heterostructure via photocurrent spectral atomic force microscopic imaging. Nano Lett. 16,
3571-3577 (2016).
27. M.-H. Doan et al., Charge transport in MoS2/WSe2 van der Waals heterostructure with
tunable inversion layer. ACS Nano 11, 3832-3840 (2017).
28. X. Sun et al., Visualizing Band Profiles of Gate-Tunable Junctions in MoS2/WSe2 Het-
erostructure Transistors. ACS Nano 15, 16314-16321 (2021).
29. H. S. Ra et al., An Asymmetry Field-Effect Phototransistor for Solving Large Exciton Bind-
ing Energy of 2D TMDCs. Adv. Mater. 34, 2107468 (2022).
30. Mini-spectrometers product lineup (Hamamatsu, 2022); https://www.
hamamatsu.com/eu/en/product/optical-sensors/spectrometers/
mini-spectrometer.html
31. H. H. Yoon et al., Strong fermi-level pinning at metal/n-Si (001) interface ensured by form-
ing an intact Schottky contact with a graphene insertion layer. Nano Lett. 17, 44-49 (2017).
32. S. Jung et al., Reduction of water-molecule-induced current-voltage hysteresis in graphene
field effect transistor with semi-dry transfer using flexible supporter. J. Appl. Phys. 125,
184302 (2019).
33. H. H. Yoon et al., Negative Fermi-level pinning effect of metal/n-GaAs (001) junction
induced by a graphene interlayer. ACS Appl. Mater. Interfaces 11, 47182-47189 (2019).
34. F. Pizzocchero et al., The hot pick-up technique for batch assembly of van der Waals het-
erostructures. Nat. Commun. 7, 11894 (2016).
15
35. R. Frisenda et al., Recent progress in the assembly of nanodevices and van der Waals het-
erostructures by deterministic placement of 2D materials.Chem. Soc. Rev. 47, 53-68 (2018).
36. K. Kinoshita et al., Dry release transfer of graphene and few-layer h-BN by utilizing ther-
moplasticity of polypropylene carbonate. npj 2D Mater. Appl. 3, 22 (2019).
37. L. Britnell et al., Field-effect tunneling transistor based on vertical graphene heterostruc-
tures. Science 335, 947-950 (2012).
38. F. Ahmed et al., Dielectric dispersion and high field response of multilayer hexagonal boron
nitride. Adv. Funct. Mater. 28, 1804235 (2018).
39. M. Faraji-Dana et al., Compact folded metasurface spectrometer. Nat. Commun. 9, 4196
(2018).
40. E. Le Coarer et al., Wavelength-scale stationary-wave integrated Fourier-transform spec-
trometry. Nat. Photonics 1, 473-478 (2007).
41. H. Sun et al., In situ formed gradient bandgap-tunable perovskite for ultrahigh-
speed color/spectrum-sensitive photodetectors via electron-donor control. Adv. Mater. 32,
1908108 (2020).
42. Y. Liu et al., Van der Waals heterostructures and devices. Nat. Rev. Mater. 1, 16042 (2016).
43. K. Novoselov, o. A. Mishchenko, o. A. Carvalho, A. Castro Neto, 2D materials and van der
Waals heterostructures. Science 353, aac9439 (2016).
44. D. Jariwala, T. J. Marks, M. C. Hersam, Mixed-dimensional van der Waals heterostructures.
Nat. Mater. 16, 170-181 (2017).
16
45. G. Konstantatos, Current status and technological prospect of photodetectors based on two-
dimensional materials. Nat. Commun. 9, 5266 (2018).
46. L. Pahun, Y. Campidelli, F. A. d’Avitaya, P. Badoz, Infrared response of Pt/Si/ErSi1.7
heterostructure: tunable internal photoemission sensor. Appl. Phys. Lett. 60, 1166-1168
(1992).
47. P. Vabbina et al., Highly sensitive wide bandwidth photodetector based on internal photoe-
mission in CVD grown p-type MoS2/graphene Schottky junction. ACS Appl. Mater. Inter-
faces 7, 15206-15213 (2015).
48. H. Yang et al., Graphene barristor, a triode device with a gate-controlled Schottky barrier.
Science 336, 1140-1143 (2012).
49. Y. J. Choi et al., Color-selective Schottky barrier modulation for optoelectric logic. ACS
Nano 14, 16036-16045 (2020).
50. H. H. Yoon et al., Tunable Quantum Tunneling through a Graphene/Bi2Se3 Heterointerface
for the Hybrid Photodetection Mechanism. ACS Appl. Mater. Interfaces 13, 58927-58935
(2021).
51. M. Du et al., Switchable photoresponse mechanisms implemented in single van der Waals
semiconductor/metal heterostructure. ACS Nano 16, 568-576 (2022).
52. X. Bai et al., Molybdenum disulfide/double-wall carbon nanotube mixed-dimensional het-
erostructures. Adv. Mater. Interfaces 9, 2200193 (2022).
53. N. Li et al., van der Waals semiconductor empowered vertical color sensor. ACS Nano 16,
8619-8629 (2022).
17
54. G. Jin et al., Heteroepitaxial van der Waals semiconductor superlattices. Nat. Nanotechnol.
16, 1092-1098 (2021).
55. W. Zhang, Q. Wang, Y. Chen, Z. Wang, A. T. Wee, Van der Waals stacked 2D layered
materials for optoelectronics. 2D Mater. 3, 022001 (2016).
56. S. Goossens et al., Broadband image sensor array based on graphene–CMOS integration.
Nat. Photonics 11, 366-371 (2017).
57. W. Liu et al., Graphene charge-injection photodetectors. Nat. Electron. 5, 281-288 (2022).
58. C. Gong et al., 2D nanomaterial arrays for electronics and optoelectronics. Adv. Funct.
Mater. 28, 1706559 (2018).
59. J. Li et al., General synthesis of two-dimensional van der Waals heterostructure arrays.
Nature 579, 368-374 (2020).
60. L. Mennel et al., Ultrafast machine vision with 2D material neural network image sensors.
Nature 579, 62-66 (2020).
61. C. Kim et al., Fermi level pinning at electrical metal contacts of monolayer molybdenum
dichalcogenides. ACS Nano 11, 1588-1596 (2017).
62. J. Jang et al., Fermi-Level Pinning-Free WSe2 Transistors via 2D van der Waals metal
contacts and their circuits. Adv. Mater. 34, 2109899 (2022).
63. X. Liu, M. S. Choi, E. Hwang, W. J. Yoo, J. Sun, Fermi level pinning dependent 2D semi-
conductor devices: challenges and prospects. Adv. Mater. 34, 2108425 (2022).
64. J. Xiao, M. Zhao, Y. Wang, X. Zhang, Excitons in atomically thin 2D semiconductors and
their applications. Nanophotonics 6, 1309-1328 (2017).
18
65. M. Massicotte et al., Dissociation of two-dimensional excitons in monolayer WSe2. Nat.
Commun. 9, 1633 (2018).
66. H. H. Yoon, W. Cai, F. Nigmatulin, Spectral Reconstruction Based on Tunable vdW
Junction Spectrometers, version 1.0, Zenodo (2022) https://doi.org/10.5281/
zenodo.7012876.
Acknowledgments
We acknowledge the provision of facilities and technical support from the Otaniemi research in-
frastructure (OtaNano-Micronova Nanofabrication Centre and OtaNano-Nanomicroscopy Cen-
tre). We thank Andreas Liapis, Mingde Du, Yunyun Dai, Suvi-Tuuli Akkanen, Juan Camilo
Arias, and Xueyin Bai for valuable discussions and Mikko Turunen, Diao Li, Yi Zhang, Vin-
cent Pelgrin, and Susobhan Das for access to the optical instruments and components. Fund-
ing: This work was supported by the Academy of Finland (Grant No. 314810, 333982, 336144,
336813, 336818, and 348920), Academy of Finland Flagship Programme (Grant No. 320167,
PREIN), the EU H2020-MSCA-RISE-872049 (IPN-Bio), EPSRC (Grant No. EP/T014601/1),
ERC (Grant No. 834742), and National Natural Science Foundation of China (Grant No.
51976122 and 52061135108). Pertti Hakonen was supported by the Jane and Aatos Erkko
foundation and the Technology Industries of Finland centennial foundation (Future Makers
2021). Author contributions: Z.S. conceived the ideas during the discussion with H.H.Y.
and F.A.. H.H.Y designed the experiments and carried out the characterizations/measurements.
H.H.Y., H.A.F., F.N., and M.G.U. fabricated the van der Waals heterostructures and spectrom-
eter devices. H.A.F. provided the home-built optical system and prepared the optical instru-
ments and components. F.A. and X.C. helped with the electrical and optoelectrical measure-
ments. H.H.Y., F.N., W.C., Z.Y., and H.C. developed the reconstruction code. W.C., Z.Y.,
19
H.C., and T.H. shared the processing strategies. H.H.Y., H.A.F., F.N., F.A., and Z.S. ana-
lyzed the data. W.C., E.D.M., P.H., K.K., H.L., and T.H. commented on the experimental
results and helped with the data analysis. H.A.F. and X.C. helped with the graphic design.
H.H.Y. wrote the manuscript, and Z.S. supervised the research. All authors participated in
the scientific discussion extensively and contributed to the manuscript writing. Correspon-
dence: addressed to H.H.Y. or Z.S.. Competing interests: The authors declare no com-
peting interests. Data and materials availability: All data needed to evaluate the conclu-
sions in the paper are present in the main text or the supplementary materials. Code used for
spectral reconstruction based on tunable vdW junction spectrometers is available at https:
//github.com/fonig/Reconstruction and archived at Zenodo (66).
Supplementary Material
Materials and Methods
Supplementary Text
Figs. S1 to S16
Table S1
References (31-66)
20