0% found this document useful (0 votes)
19 views

Single Photon Frequency Conversion For Frequency Multiplexed Quantum Networks in The Telecom Band

1) The document describes an experiment demonstrating efficient single photon frequency conversion between wavelength division multiplexing channels in the telecom band using a cascaded second-order nonlinear interaction. 2) The experiment achieved a maximum conversion efficiency of 55% and a noise subtracted Hong-Ou-Mandel dip visibility of 84.5% between converted photons. 3) This integrated quantum frequency conversion protocol can interface a broad spectrum of frequencies compatible with quantum memories and fiber networks, enabling long-range high-speed quantum communication using frequency multiplexing.

Uploaded by

Stephen Yin
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
19 views

Single Photon Frequency Conversion For Frequency Multiplexed Quantum Networks in The Telecom Band

1) The document describes an experiment demonstrating efficient single photon frequency conversion between wavelength division multiplexing channels in the telecom band using a cascaded second-order nonlinear interaction. 2) The experiment achieved a maximum conversion efficiency of 55% and a noise subtracted Hong-Ou-Mandel dip visibility of 84.5% between converted photons. 3) This integrated quantum frequency conversion protocol can interface a broad spectrum of frequencies compatible with quantum memories and fiber networks, enabling long-range high-speed quantum communication using frequency multiplexing.

Uploaded by

Stephen Yin
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 5

Single photon frequency conversion for frequency multiplexed quantum networks in

the telecom band

Paul Fisher,1, ∗ Robert Cernansky,1 Ben Haylock,1 and Mirko Lobino1, 2, †


1
Centre for Quantum Computation and Communication Technology (Australian Research Council),
Centre for Quantum Dynamics, Griffith University, Brisbane, QLD 4111, Australia
2
Queensland Micro- and Nanotechnology Centre, Griffith University, Brisbane, QLD 4111, Australia
High-speed long-range quantum communication requires combining frequency multiplexed pho-
tonic channels with quantum memories. We experimentally demonstrate an integrated quantum
frequency conversion protocol that can convert between wavelength division multiplexing channels
in the telecom range with an efficiency of 55 ± 8% and a noise subtracted HOM dip visibility of
84.5%. This protocol is based on a cascaded second order nonlinear interaction and can be used to
interface a broad spectrum of frequencies with narrowband quantum memories, or alternatively as
arXiv:2104.01881v1 [quant-ph] 5 Apr 2021

a quantum optical transponder, efficiently interfacing different regions of a frequency-multiplexed


spectrum.

Optical frequency conversion [1] is an enabling technol- with a single pump conversion scheme, this is usually lim-
ogy for quantum computation [2] and quantum commu- ited in bandwidth because of the narrow phase-matching
nication [3]. It plays a crucial role in interfacing quan- condition. Fortunately, it is possible to emulate the two
tum memories, which usually work outside of the tele- pump χ(3) scheme by cascading two χ(2) frequency con-
com spectrum (around 1550nm), with the current optical version processes. This involves performing two concur-
fiber infrastructure. Quantum memories [4] are an inte- rent frequency conversions between distant wavelengths
gral part of the quantum repeater protocol and they can to achieve a small frequency shift. Originally suggested
be implemented in platforms such as trapped ions [5], for tuneable conversion of classical signals [22], it was
atomic ensembles [6], and rare-earth doped crystals [7]. later demonstrated how a single waveguide device can
Quantum frequency conversion has been used to demon- perform frequency conversion between any pair of chan-
strate remote entanglement [8] between memories sepa- nels in the WDM grid with an average efficiency >80%
rated by up to 50 km of optical fibre [9], and for the [23].
conversion of single photons emitted from trapped ions Here, we demonstrate the scheme presented in [23] for
[10, 11] into the 1550 nm region. high efficiency frequency shifting between WDM channels
Recent developments have shown potential light stor- on single photons. This protocol has the dual advantages
age at telecom wavelengths using erbium doped materi- of two orders of magnitude higher conversion efficiency,
als [12, 13] and engineered optomechanical systems [14]. and tuneability across the full telecom bandwidth when
While compatible with the fiber network, these memo- compared with similar, simplified techniques [24]. We
ries have a storage bandwidth of the order of hundreds of achieve a conversion efficiency of up to 55 ± 8% and mea-
MHz, which is only a small fraction of the 7.2 THz avail- sured quantum interference between converted photons
able for the 72 channels in the wavelength division mul- with a HOM dip visibility of 49.3% (84.5% when noise is
tiplexing (WDM) grid. This is why dynamic frequency subtracted). We also present solutions to current limita-
conversion between different WDM channels plays a key tions of our set-up and for future practical applications.
role if we want exploit the full potential of the quantum Our frequency conversion scheme is illustrated in
fiber network via frequency multiplexing. Fig. 1(a) and consists of concurrent stages of sum fre-
quency generation (SFG) and difference frequency gen-
Conversion between WDM channels can be achieved in
second and third order nonlinear materials. Third order eration (DFG) in a nonlinear waveguide. In the SFG
nonlinear materials use a two pump scheme [15] with ex- stage, the photon at the signal frequency is upconverted
periments performed in nonlinear optical fibres [16, 17], to the intermediate frequency by pump 1, then in the
and in silicon microresonators [18]. The relative weak- DFG stage pump 2 downconverts the photon from the
intermediate frequency to the target frequency. The dif-
ness of the χ(3) nonlinear susceptibility means that these
ference in phase velocities between the interacting modes
schemes require powerful pumps, very long nonlinear fi-
results in the phase mismatches
bres or high quality resonators. Each of these present
limits on the flexibility and bandwidth of the frequency ∆kSF G = kSF G − ks − kP 1 (1)
conversion.
∆kDF G = kSF G − kt − kP 2 (2)
Second order nonlinear (χ(2) ) materials are usually
more efficient for frequency conversion, with experiments for the SFG and DFG processes (see Fig 1(b)), where
performed on periodically poled waveguides in lithium kx indicates the wavevector of a mode in the waveguide
niobate [19–21]. Though χ(2) materials are compatible and subscript stands for s for signal, t for target, and P1
2

SPDC

ωs

FIG. 1. (a) Interaction between the frequency components (ω). Frequency shift ∆ω = ωt − ωs = ωP1 − ωP2 . (b) The difference
in wavenumber (k) between all the frequency components creates two phase mismatches (∆kSF G and ∆kDF G ) for the two
cascaded processes. Includes the average phase mismatch K and their difference δK . (c) Schematic layout of the experiment.
Most parts realised using bulk optical components. Beam splitter (BS), Erbium Doped Fibre Amplifier (EDFA), Band Pass
Filter (BPF), Grating Filter (GF), Avalanche Photodiodes (APDs), Time to Digital Converter (TDC), continuous wave pump
lasers (CW Pumps).

and P2 for the two pumps. These two phase mismatches conversion efficiency,
can be partially compensated by quasi-phase-matching
hâ†t (L)ât (L)i 16Q2
 q 
in materials that are compatible with periodic poling. L
η= = sin4 δK
2
+ 4Q ,
hâ†s (0)âs (0)i
2
It was shown in [23] that the optimal poling period for (δK + 4Q)2 4
this cascaded interaction is the one that compensates for (5)
2
the average phase mismatch K = (∆kDF G + ∆kSF G )/2, for a waveguide of length L where Q = 2X12 |E1 | =
while the difference δK = ∆kDF G − ∆kSF G determines 2
2X22 |E2 | . The maximum possible conversion efficiency
the limit to the overall conversion efficiency. q
2
To describe the single photon frequency conversion, we obtained when L/4 δK + 4Q = π/2 + mπ and is given
extend the typical formulation of quantum frequency con- by,
version [1, 15], where pumps are treated as classical fields !2
2 2
and phase mismatches are included using a rotating wave δK L
ηmax = 1− (6)
approximation, and begin with a Hamiltonian, 4π 2

H = X1 E1 âs â†SFG +X1 E1∗ â†s âSFG + X2 E2 ât â†SFG + for m = 0 where the condition δK 2 2
L ≤ 4π 2 (1 + 2m)2 has
δ δ to be satisfied for all m.
X2 E2∗ â†t âSFG − K â†s âs + Kâ†SFG âSFG + K â†t ât .
2 2 The χ(2) nonlinearity, present in non-centrosymetric
(3) materials like lithium niobate, is many orders of mag-
nitude larger than the χ(3) nonlinearity. Consequently,
X1 and X2 are the coupling strengths of the SFG and
shorter devices using weaker pumps can achieve the same
DFG processes, E1 and E2 are the electric field strengths
conversion efficiency as several metres of nonlinear fibre
of pump 1 and pump 2, and â†s,SFG,t and âs,SFG,t are the
using powerful pumps. Comparing our 3.8 cm waveguide
creation and annihilation operators for the signal, sum
to a nonlinear fibre 100 m long [17], we estimate max-
frequency and target modes. Time evolution inside a
imum conversion with 1.2 W of total pump power (see
waveguide is equivalent to spatial propagation, giving us
Fig. 2) while the fibre requires 20 W. Maximum conver-
a set of spatial Heisenberg equations of motion for the
sion can be reached with less power in a much longer
annihilation operators,
nonlinear fibre (300 mW in 750 m [16]), but as Eq. 6
dâs δ shows, extra length is detrimental to maximum conver-
= −i K âs + iX1 E1∗ âSFG , (4a) sion efficiency. This is also true since the phase mismatch
dz 2
dâSFG difference δK is the same as the phase mismatch for the
= iX1 E1 âs + iKâSFG + iX2 E2 ât , (4b) χ(3) process, with similar effect to maximum efficiency
dz
dât δ shown in Eq. 6.
= iX2 E2∗ âSFG + i K ât . (4c) A schematic of the experimental setup is shown in
dz 2
Fig. 1(c). The photon pairs were generated by SPDC
Under the conditions that the poling period is given by in a reverse-proton-exchanged periodically poled lithium
Λ = 2π/K and the pump powers are tuned so that niobate (RPE:PPLN) waveguide [25] using a erbium fem-
2 2
X12 |E1 | = X22 |E2 | , we can solve Eqs. 4 and find the tosecond fiber laser with a repetition rate of 25 MHz.
3

ond harmonic generation of 46 %W−1 cm−2 . It was kept


100
at a temperature of 122.5 ◦ C for optimal phase match-
ing and to avoid high-power photorefractive effects. The
80 facets of the chip were anti-reflection coated with a re-
Efficiency (%)

flectivity less than 1%.


60 After conversion the single photons were separated
from the pumps by a band-pass filter (Semrock 1550/3)
40 with the reflected pumps used to monitor the coupled
power. A diffraction grating (Thorlabs 1200/mm holo-
20 graphic) was used to further suppress the amount of
pump light present in the converted photon spatial mode.
0 The grating was operated close to the Littrow angle,
0 500 1000 sending converted photons back through the bandpass
Total Pump (mW) filter to a mirror, then through the filter again. The
transmisson spectrum of this filtering stage was 28.6 GHz
FIG. 2. Conversion efficiency for three different pump pow- FWHM centred at 1535.82 nm. A similar filter set-up was
ers measured as the difference between counts with photons incorporated into the delay stage to maximise the spec-
blocked and unblocked. Normalised by count rate with no tral overlap between converted and unconverted photons.
conversion. Solid line is the theoretical efficiency with maxi- The single photons were detected using two InGaAs
mum at P=1.2 W.
single photon detectors (SPD, IDQuantique ID200 and
ID201) gated with a 2.5 ns window at 1 MHz rate and
synchronized with the laser pulses. Coincidences were
A 100 GHz portion of the laser spectrum, centred at counted using a time to digital converter (UQDevices
1537.40 nm (corresponding to channel 50 of the dense Logic16). Figure 2 shows the conversion efficiency as a
WDM grid), was selected by a WDM module, ampli- function of the total pump power which reaches 55 ± 8%
fied, frequency doubled, and used to generate spectrally for a total pump power of 600 mW. The efficiency was
separated photon pairs around 1537.40 nm. The aver- measured as the ratio between single counts of converted
age second harmonic power was 20 µW, corresponding photons by the SPD, versus single counts of unconverted
to a pair generation probability per pulse of ∼1-2% and photons with the pumps off and the filters tuned to
a coincidence to accidental ratio of 8.55. The photons 1538.98 nm. Noise counts from the pumps were sub-
were separated in a WDM module, with signal photons tracted by blocking the single photons before the waveg-
at 1538.98 nm (channel 48) sent to be converted and idler uide. The theoretical fit in Fig. 2 shows that our waveg-
photons at 1535.82 nm (channel 52) sent to a delay stage. uide requires a total pump power of approximately 1.2
Both before and after the SPDC more than 150 dB of fil- W for a theoretical maximum conversion of 89.6%. The
tering was used to ensure clean photon pair generation. fit includes the effects of waveguide propagation loss and
The pumps used for the frequency conversion of the the wrong pump interaction.
signal photons were two tunable CW diode lasers with The high efficiency χ(2) process allows undesirable in-
pump 1 set to 1550.28 nm and pump 2 set to 1553.49 nm, teractions between pumps and signal, primarily intro-
400 GHz lower in frequency to make the +400 GHz shift ducing an effective loss through signal photons interact-
in the converted light to match the frequency of the tar- ing with the wrong pump. The pumps themselves may
get single photon. Both lasers were passed thru a sin- also generate second harmonics or undergo SFG between
gle erbium-doped fibre amplifier and their output powers themselves, and these new frequencies can create noise
balanced to within 0.2% for optimal conversion. Follow- photons by SPDC. Although these interactions are al-
ing amplification to a combined power of 16 W, these ways present, their effects can be reduced by choosing
pumps passed through approximately 90 dB of bandpass pumps further away from the signal and target frequency
filtering to eliminate any amplifier noise near the photon as described in [23].
frequencies. Using polarization optics, the pumps were The primary source of noise is Raman scattering. This
then combined with the signal photons from SPDC at a is supported by the linear increase in noise with input
ratio of about 96:4, favouring the single photons. This power shown in Fig. 3a. Since the total amount of com-
left approximately 600 mW of pump power for conversion bined second harmonic and sum frequency generated by
and allowed the pump power to be tuned using a half- 500 mW of pump is approximately 2 mW, these process
wave plate, with only a small penalty to photon losses. and any subsequent SPDC are poorly phase matched and
The combined pumps and single photons were then very inefficient. Figure 3b shows the noise spectrum near
coupled to another RPE:PPLN waveguide for frequency our possible choices of pumps frequencies. Considering
conversion. The frequency conversion device was 3.8 cm that the pumps have to be 400 GHz apart, only marginal
long and had a normalized conversion efficiency for sec- improvement can be achieved tuning their frequencies.
4

2.5 Uncorrelated Raman photons are the main limiting


Counts factor of our device. Several approaches can be used to
(kHz) 2.0 mitigate this problem including using pump wavelengths
1.5
around 2.4 µm, which also allow for efficient conversion
[23]. This is because the telecom region lies on the oppo-
350 400 450 500 550 600 650 site side of an inflection in the refractive index curve of
Total Pump (mW) lithium niobate to 2.4 µm. Consequentially, this allows
the difference in phase mismatches δK to be even smaller
2.0 than what is allowed with telecom pumps, improving ef-
Counts

ωP2
(kHz)

1.0
ωP1 ficiency. Moving to picosecond pulsed pumps with better
time gating of the detectors and heralded generation of
0 the single photons can also be implemented to reduce the
0 100 200 300 400 500 600 700
noise by at least two orders of magnitude.
Freq. (+192800 GHz)
In conclusion, we have demonstrated a scheme for
FIG. 3. (a) Noise counts for the total pump power P1 + P2 . the conversion of single photons between telecom WDM
(b) Noise spectrum from the conversion waveguide with a channels using cascaded sum and difference frequency
single pump at 500 mW. Vertical bars indicate chosen pump generation. Using our waveguide, we have achieved a
frequencies which minimised combined noise. conversion efficiency up to 55 ± 8% and shown quantum
interference between converted photons. Although Ra-
man generation in the waveguide reduced the visibility
6 of the HOM dip, experimental improvements around
Coincidence Rate (Hz)

the pump lasers can be used to mitigate this effect.


This scheme will allow frequency mulitplexed quantum
communication across the entire telecom spectrum to
4
be efficiently interfaced with narrowband rare-earth
quantum memories, potentially enabling long distance
high bandwidth quantum communication.
2
This work was supported by the Australian Re-
search Council (ARC) Centre of Excellence for Quan-
0 tum Computation and Communication Technology
-30 -20 -10 0 10 20 30
(CE170100012), and the Griffith University Research
Delay (ps) Infrastructure Program. ML was supported by the
Australian Research Council (ARC) Future Fellowship
FIG. 4. Blue curve is the Hong-Ou-Mandel interference dip (FT180100055). PF was supported by the Australian
between converted and unconverted photons. Red line is the Government Research Training Program Scholarship.
constant noise floor measured as the combined rate of acci- BH was supported by the Griffith University Postdoc-
dental coincidences between noise only, unconverted only and
toral Fellowship. This work was performed in part at
noise, and converted only and noise.
the Queensland node of the Australian National Fabrica-
tion Facility, a company established under the National
Collaborative Research Infrastructure Strategy to pro-
Finite temperature tuning range of our phase matching
vide nano- and microfabrication facilities for Australia’s
limited operation to this region.
researchers.
Indistinguishability between the converted and the un-
converted photon was verified with a Hong-Ou-Mandel
(HOM) interference [26] measurement in a fibre beam
splitter, converting the photons with a total pump power

of 500 mW. A tuneable delay was added to the uncon- [email protected]

verted (target) channel and the number of coincidences [email protected]
after the beamsplitter was measured as a function of de- [1] P. Kumar, Quantum frequency conversion, Optics Let-
lay . Figure 4 shows the resulting HOM dip with a raw ters 15, 1476 (1990).
[2] N. K. Langford, S. Ramelow, R. Prevedel, W. J. Munro,
visibility of 49.3 ± 2.7%, limited by the noise from un-
G. J. Milburn, and A. Zeilinger, Efficient quantum com-
correlated Raman photons. Noise subtracted visibility is puting using coherent photon conversion, Nature 478,
84.5 ± 4.6% which is limited by the imperfect 49.3:50.7 360 (2011).
splitting ratio of the fiber splitter and the spectral over- [3] N. Gisin and R. Thew, Quantum communication, Nature
lap between the two filters of 96%. Photonics 1, 165 (2007).
5

[4] A. I. Lvovsky, B. C. Sanders, and W. Tittel, Optical Raymer, Translation of quantum states by four-wave
quantum memory, Nature Photonics 3, 706 (2009). mixing in fibers, Optics Express 13, 9131 (2005).
[5] Y. Wang, M. Um, J. Zhang, S. An, M. Lyu, J.-N. Zhang, [16] A. S. Clark, S. Shahnia, M. J. Collins, C. Xiong, and
L.-M. Duan, D. Yum, and K. Kim, Single-qubit quan- B. J. Eggleton, High-efficiency frequency conversion in
tum memory exceeding ten-minute coherence time, Na- the single-photon regime, Optics Letters 38, 947 (2013).
ture Photonics 11, 646 (2017). [17] C. Joshi, A. Farsi, S. Clemmen, S. Ramelow, and A. L.
[6] D. J. Saunders, J. H. D. Munns, T. F. M. Champion, Gaeta, Frequency multiplexing for quasi-deterministic
C. Qiu, K. T. Kaczmarek, E. Poem, P. M. Ledingham, heralded single-photon sources, Nature Communications
I. A. Walmsley, and J. Nunn, Cavity-Enhanced Room- 9, 847 (2018).
Temperature Broadband Raman Memory, Phys. Rev. [18] Q. Li, M. Davanço, and K. Srinivasan, Efficient and low-
Lett. 116, 90501 (2016). noise single-photon-level frequency conversion interfaces
[7] M. P. Hedges, J. J. Longdell, Y. Li, and M. J. Sellars, using silicon nanophotonics, Nature Photonics 10, 406
Efficient quantum memory for light, Nature 465, 1052 (2016).
(2010). [19] B. Albrecht, P. Farrera, X. Fernandez-Gonzalvo, M. Cris-
[8] L.-M. Duan, M. D. Lukin, J. I. Cirac, and P. Zoller, Long- tiani, and H. de Riedmatten, A waveguide frequency con-
distance quantum communication with atomic ensembles verter connecting rubidium-based quantum memories to
and linear optics, Nature 414, 413 (2001). the telecom C-band, Nature Communications 5, 3376
[9] Y. Yu, F. Ma, X.-Y. Luo, B. Jing, P.-F. Sun, R.-Z. (2014).
Fang, C.-W. Yang, H. Liu, M.-Y. Zheng, X.-P. Xie, W.-J. [20] A. Lenhard, J. Brito, M. Bock, C. Becher, and J. Eschner,
Zhang, L.-X. You, Z. Wang, T.-Y. Chen, Q. Zhang, X.-H. Coherence and entanglement preservation of frequency-
Bao, and J.-W. Pan, Entanglement of two quantum mem- converted heralded single photons, Optics Express 25,
ories via fibres over dozens of kilometres, Nature 578, 240 11187 (2017).
(2020). [21] S. Kasture, F. Lenzini, B. Haylock, A. Boes, A. Mitchell,
[10] M. Bock, P. Eich, S. Kucera, M. Kreis, A. Lenhard, E. W. Streed, and M. Lobino, Frequency conversion be-
C. Becher, and J. Eschner, High-fidelity entanglement be- tween UV and telecom wavelengths in a lithium nio-
tween a trapped ion and a telecom photon via quantum bate waveguide for quantum communication with Yb +
frequency conversion, Nature Communications 9, 1998 trapped ions, Journal of Optics 18, 104007 (2016).
(2018). [22] Y. Jun, A. Takada, E. Yamazaki, O. Tadanaga,
[11] T. Walker, K. Miyanishi, R. Ikuta, H. Takahashi, H. Miyazawa, and M. Asobe, Selective wavelength con-
S. Vartabi Kashanian, Y. Tsujimoto, K. Hayasaka, T. Ya- version using PPLN waveguide with two pump configura-
mamoto, N. Imoto, and M. Keller, Long-distance single tion, in Conference on Lasers and Electro-Optics (2003)
photon transmission from a trapped ion via quantum fre- pp. 1135–1136.
quency conversion, Phys. Rev. Lett. 120, 203601 (2018). [23] P. Fisher, M. Villa, F. Lenzini, and M. Lobino, Integrated
[12] E. Saglamyurek, J. Jin, V. B. Verma, M. D. Shaw, Optical Device for Frequency Conversion Across the Full
F. Marsili, S. W. Nam, D. Oblak, and W. Tittel, Quan- Telecom C-Band Spectrum, Physical Review Applied 13,
tum storage of entangled telecom-wavelength photons in 24017 (2020).
an erbium-doped optical fibre, Nature Photonics 9, 83 [24] T. Xiang, Q.-C. Sun, Y. Li, Y. Zheng, and
(2015). X. Chen, Single-photon frequency conversion via cas-
[13] M. Rančić, M. P. Hedges, R. L. Ahlefeldt, and M. J. caded quadratic nonlinear processes, Physical Review A
Sellars, Coherence time of over a second in a telecom- 97, 063810 (2018).
compatible quantum memory storage material, Nature [25] F. Lenzini, S. Kasture, B. Haylock, and M. Lobino,
Physics 14, 50 (2017). Anisotropic model for the fabrication of annealed and re-
[14] A. Wallucks, I. Marinković, B. Hensen, R. Stockill, and verse proton exchanged waveguides in congruent lithium
S. Gröblacher, A quantum memory at telecom wave- niobate, Optics Express 23, 1748 (2015).
lengths, Nature Physics 16, 772 (2020). [26] C. K. Hong, Z. Y. Ou, and L. Mandel, Measurement of
[15] C. J. McKinstrie, J. D. Harvey, S. Radic, and M. G. subpicosecond time intervals between two photons by in-
terference, Phys. Rev. Lett. 59, 2044 (1987).

You might also like