---
title: 'Erbium-Doped TFLN: Active Telecom Photonics'
url: https://www.emergentmind.com/topics/erbium-doped-thin-film-lithium-niobate
type: topic
---

# Erbium-Doped TFLN: Active Telecom Photonics

Erbium-doped thin-film lithium niobate, usually denoted Er:TFLN or Er:LNOI, is a lithium-niobate-on-insulator material system in which $\mathrm{Er}^{3+}$ provides telecom-band gain while the lithium niobate host contributes strong electro-optic, piezoelectric, and second-order nonlinear functionality. In the reported literature, this combination supports integrated waveguide amplifiers, microring and microdisk lasers, wavelength-routing components, and cryogenic light–matter interfaces operating on the native $\mathrm{Er}^{3+}$ transition near $1.53\text{–}1.56\,\mu\mathrm{m}$, with pumping implemented either near $974\text{–}980\,\mathrm{nm}$ through the ${}^4I_{15/2}\rightarrow{}^4I_{11/2}$ transition or near $1480\,\mathrm{nm}$ through direct excitation of ${}^4I_{13/2}$ [2101.06994] [2108.09740] [2605.14777].

## 1. Platform, material system, and erbium incorporation

Reported Er:TFLN platforms span several crystallographic cuts and stack geometries. Z-cut devices include $600\,\mathrm{nm}$ Er-doped LiNbO$_3$ films on $2\,\mu\mathrm{m}$ SiO$_2$ with either $400\,\mu\mathrm{m}$ Si or $0.5\,\mathrm{mm}$ Si handles, $700\,\mathrm{nm}$ Er:LN films on $2\,\mu\mathrm{m}$ SiO$_2$ on $500\,\mu\mathrm{m}$ LN handles, and $500\,\mathrm{nm}$ Er:TFLN on $4.7\,\mu\mathrm{m}$ buried SiO$_2$ on $500\,\mu\mathrm{m}$ Si. X-cut and x-cut implementations include $300\,\mathrm{nm}$ and $500\,\mathrm{nm}$ films, including symmetric SiO$_2$-clad quantum-memory devices and undoped x-cut films subsequently implanted with erbium [2108.09740] [2101.00783] [2508.11941] [2511.04442] [2605.14777].

Erbium incorporation has been realized by several routes. Many amplifier and laser papers use bulk-doped crystals followed by ion slicing or smart-cut transfer, producing films with nominal erbium concentrations of $0.5\text{–}1\,\mathrm{mol}\%$ and effectively uniform dopant distribution through the thin film. In a traveling-wave amplifier, a uniform $1\,\mathrm{mol}\%$ film was explicitly associated with an approximate density of $1.9\times10^{20}\,\mathrm{cm}^{-3}$ as an order-of-magnitude estimate; another amplifier reported $N_{\mathrm{ET}} = 0.72\times10^{20}\,\mathrm{cm}^{-3}$ from XPS/TOF-SIMS and a fluorescence lifetime $T_{\mathrm{pl}}\approx 2.3\,\mathrm{ms}$ [2101.06994] [2101.00783] [2108.08044]. Ion implantation has been used both for post-fabrication ensemble integration and for deterministic placement. One study implanted erbium at $350\,\mathrm{keV}$ through a $30\,\mathrm{nm}$ SiO$_2$ cap into a $600\,\mathrm{nm}$ z-cut LNOI film, with a depth distribution peaking near $80\,\mathrm{nm}$ below the top surface and average depth near $50\,\mathrm{nm}$ inside LN; a later study used focused-ion-beam implantation of isotopically selected $^{170}$Er at $75\,\mathrm{keV}$, achieving a projected range of $25.4\,\mathrm{nm}$ and sub-$100\,\mathrm{nm}$ lateral placement precision [1912.07584] [2511.04442]. At the cryogenic end, isotopically purified $^{167}\mathrm{Er}^{3+}$ with nominal isotopic purity $95.6\%$ has been incorporated into TFLN microrings at $500\,\mathrm{ppm}$ and $50\,\mathrm{ppm}$, enabling hyperfine-based quantum-memory protocols [2605.14777].

The operative optical transition is the native telecom-band ${}^4I_{15/2}\leftrightarrow{}^4I_{13/2}$ manifold. Pumping near $974\text{–}980\,\mathrm{nm}$ excites ${}^4I_{15/2}\rightarrow{}^4I_{11/2}$, followed by nonradiative relaxation to ${}^4I_{13/2}$, whereas $1480\,\mathrm{nm}$ pumping addresses ${}^4I_{13/2}$ directly. Strong green upconversion fluorescence under near-$980\,\mathrm{nm}$ excitation is repeatedly reported and is used both as evidence of erbium activation and as an indicator of upconversion-related processes at elevated excitation densities [2105.00214] [2108.09740] [2101.06994] [2312.15601]. The host’s electro-optic response is central to the platform identity. For z-cut LN with suitable polarization alignment, the reported first-order relation is
$$
\Delta n \approx -\frac{1}{2} n^3 r_{33} E,
$$
with $r_{33}=30.8\times10^{-12}\,\mathrm{m/V}$ cited for LiNbO$_3$, linking Er-based gain with electrically reconfigurable photonics on the same substrate [2108.09740].

## 2. Waveguide amplifiers and gain engineering

Integrated Er:TFLN amplification progressed rapidly from millimeter-scale proof-of-principle devices to centimeter-scale, externally useful amplifiers. A compact $5\,\mathrm{mm}$ Er:LNOI waveguide amplifier pumped at $974.5\,\mathrm{nm}$ demonstrated on-chip net gain $>5\,\mathrm{dB}$ for a $1530\,\mathrm{nm}$ signal with $21\,\mathrm{mW}$ pump power; the same work reported a conservative passive loss of $\sim2\,\mathrm{dB/cm}$ at $1530\,\mathrm{nm}$ and fiber-to-fiber insertion loss of $\approx12\,\mathrm{dB}$ for the $5\,\mathrm{mm}$ device [2101.06994]. Shortly afterward, a $3.6\,\mathrm{cm}$ spiral amplifier fabricated by PLACE achieved a maximum internal net gain of $18\,\mathrm{dB}$ at $1530\,\mathrm{nm}$, corresponding to about $5\,\mathrm{dB/cm}$, with $3\,\mathrm{dB}$ gain bandwidth $>40\,\mathrm{nm}$ centered around $\sim1540\,\mathrm{nm}$ and signal-band propagation loss $\alpha_s\approx0.16\,\mathrm{dB/cm}$ extracted from a microring with loaded $Q\approx2.6\times10^6$ [2101.00783]. A $2.58\,\mathrm{cm}$ amplifier provided a more measurement-complete characterization: $27.94\,\mathrm{dB}$ signal enhancement, $16.0\,\mathrm{dB}$ internal net gain, $6.20\,\mathrm{dB/cm}$ gain density, $-8.84\,\mathrm{dBm}$ saturation power, $4.59\,\mathrm{dB/mW}$ power conversion efficiency, and $4.49\,\mathrm{dB}$ noise figure at $1531.6\,\mathrm{nm}$ under bidirectional $1484\,\mathrm{nm}$ pumping [2108.08044].

The reported gain model is the standard erbium traveling-wave expression
$$
g(\lambda)=\Gamma\left[\sigma_e(\lambda)N_2-\sigma_a(\lambda)N_1\right]-\alpha,
$$
with $\Gamma$ the mode–dopant overlap, $\sigma_e$ and $\sigma_a$ the emission and absorption cross-sections, $N_1+N_2=N$, and $\alpha$ the passive loss [2105.00214] [2405.16583]. Within this framework, several design directions emerged. Spiral amplifiers on $1\,\mathrm{mol}\%$ Er:LNOI reached a maximum internal net gain of $8.3\,\mathrm{dB}$ at $1530\,\mathrm{nm}$, or $\approx15.6\,\mathrm{dB/cm}$, in a compact $5.3\,\mathrm{mm}$ device with $\sim0.06\,\mathrm{mm}^2$ footprint, while showing net internal gain $>3\,\mathrm{dB}$ across $\approx1530\text{–}1570\,\mathrm{nm}$ [2105.00214]. A Ta$_2$O$_5$-cladded Er:LN amplifier extended the interaction length to $\approx10\,\mathrm{cm}$ and exceeded $20\,\mathrm{dB}$ internal net gain at $1532\,\mathrm{nm}$ with $\approx55\,\mathrm{mW}$ on-chip bidirectional $976\,\mathrm{nm}$ pump, the reported explanation being that the high-index cladding reduced core confinement enough to mitigate quenched-ion absorption while maintaining useful overlap with the doped film [2111.05571].

Power scaling later required explicit management of excited-state absorption, cooperative upconversion, and thermal load. A segmented large-mode-area architecture used $9\,\mu\mathrm{m}$-wide gain sections, $1\,\mu\mathrm{m}$-wide bends, and $1\,\mathrm{mm}$ adiabatic tapers over $\approx7\,\mathrm{cm}$ total length in a $1\times1\,\mathrm{cm}^2$ footprint. Under bidirectional $1480\,\mathrm{nm}$ pumping, it achieved on-chip output power up to $\approx113\,\mathrm{mW}$, on-chip net gain of about $16\,\mathrm{dB}$, small-signal gain up to $\approx20.5\,\mathrm{dB}$ on chip, and $\approx52\%$ power-conversion efficiency; the measured noise figure was $\approx7.1\,\mathrm{dB}$ under conditions corresponding to $\approx16\,\mathrm{dB}$ on-chip gain [2405.16583]. A later monolithic amplifier incorporated edge couplers on a $2\,\mathrm{mm}\times25\,\mathrm{mm}$ chip and reported $>18\,\mathrm{dB}$ fiber-to-fiber net gain with bidirectional $\sim1480\,\mathrm{nm}$ pumping, fiber-to-fiber noise figures around $5\,\mathrm{dB}$, and output powers above $13\,\mathrm{dBm}$ at the output fiber, indicating that coupling-loss reduction had become as important as internal gain itself [2508.11941].

Across these demonstrations, two trends recur. First, $1480\,\mathrm{nm}$ pumping is repeatedly favored when mode overlap, quantum defect, or noise figure dominate system performance [2108.08044] [2401.05970] [2508.11941]. Second, amplifier figures of merit are set jointly by active-ion physics and photonic geometry: spiral compactness, cladding-engineered confinement, large-mode-area sections, and low-loss edge coupling are all used to move beyond internal-net-gain demonstrations toward deployable external gain.

## 3. Laser architectures on Er:TFLN

Er:TFLN lasers have been implemented in whispering-gallery, Fabry–Pérot, photonic-crystal, and hybrid-coupled forms. A coupled microdisk–microring “photonic molecule” on a $1\%\,\mathrm{mol}$ Z-cut Er:LNOI wafer demonstrated an integrated tunable whispering-gallery single-mode laser near $1560\,\mathrm{nm}$: the baseline device produced lasing at $1560.40\,\mathrm{nm}$ with threshold pump power $1.31\,\mathrm{mW}$, slope efficiency $4.41\times10^{-5}$, peak output power $-44.02\,\mathrm{dBm}$ at $2.13\,\mathrm{mW}$ pump, and side-mode suppression ratio of $31.4\,\mathrm{dB}$ [2108.09740]. The single-mode mechanism was explicitly attributed to Vernier-like super-mode selection between resonators with distinct free spectral ranges, using the standard whispering-gallery condition
$$
m\lambda = 2\pi n_{\mathrm{eff}}R
$$
and an effective photonic-molecule spacing reported as $\approx13.7\,\mathrm{nm}$ near $1560\,\mathrm{nm}$ [2108.09740].

Other resonant geometries emphasized monolithic coupling and tunability. An electro-optically tunable microring laser on Er:LNOI used a vertically coupled undoped-LN bus waveguide, reached a measured resonator $Q$ of $1.54\times10^5$ at $1542\,\mathrm{nm}$, showed lasing threshold below $3.5\,\mathrm{mW}$ under $980\,\mathrm{nm}$ pumping, and achieved an electro-optic tuning coefficient of $0.33\,\mathrm{pm/V}$ with a continuous tuning range of $0.2\,\mathrm{nm}$ [2103.08197]. A Fabry–Pérot laser defined by Sagnac loop reflectors had a footprint of $6.5\,\mathrm{mm}\times1.5\,\mathrm{mm}$, loaded $Q$ factor $1.6\times10^5$, free spectral range $63\,\mathrm{pm}$, threshold $\approx6\,\mathrm{mW}$, maximum total output power $44.7\,\mu\mathrm{W}$, and slope efficiency $\approx0.18\%$ while maintaining single-mode lasing around $1550\,\mathrm{nm}$ with SMSR $>25\,\mathrm{dB}$ [2212.12698]. A PLACE-fabricated, waveguide-coupled microring laser operating near $1530.9\,\mathrm{nm}$ used a ring of perimeter $\approx1\,\mathrm{mm}$, exhibited loaded $Q=3.2\times10^5$, free spectral range $\approx1\,\mathrm{nm}$, and threshold $\approx24.5\,\mathrm{mW}$, with single-frequency behavior attributed to alignment between the sharp erbium gain peak and one ring resonance [2201.11959]. Electrical pump integration has also been demonstrated by butt-coupling a commercial $980\,\mathrm{nm}$ diode chip to an Er:TFLN microring chip, producing single-mode lasing at $1531\,\mathrm{nm}$ in a $3\,\mathrm{mm}\times4\,\mathrm{mm}\times0.5\,\mathrm{mm}$ hybrid package, with optical threshold $6\,\mathrm{mW}$, threshold current $0.5\,\mathrm{A}$ at $1.64\,\mathrm{V}$, and measured linewidth $0.05\,\mathrm{nm}$ [2209.01533].

Two device classes pushed threshold or output power in more specialized directions. Er–Yb co-doped TFLN microdisks operated in coherent polygon modes, giving loaded $Q$ factors $4.79\times10^5$ at $974.79\,\mathrm{nm}$ and $1.28\times10^5$ at $1531.40\,\mathrm{nm}$, an on-chip threshold $P_{\mathrm{th}}\approx1.04\,\mu\mathrm{W}$, external conversion efficiency $4.06\times10^{-3}$, and SMSR $23.4\,\mathrm{dB}$ under $980\,\mathrm{nm}$-band pumping [2309.10512]. At the opposite end of cavity volume, a heterogeneous photonic-crystal nanobeam cavity on $1\,\mathrm{mol}\%$ Er:TFLN achieved single-mode lasing at $1559.63\,\mathrm{nm}$ with $Q=1.2\times10^4$, effective mode volume $1.44(\lambda/n)^3\approx0.645\,\mu\mathrm{m}^3$, threshold power $163\,\mu\mathrm{W}$, and strong pump-induced photorefractive blueshifting [2312.15601]. These results indicate that Er:TFLN laser design does not converge to a single canonical cavity type; instead, the platform has supported threshold minimization, electrical or EO tunability, single-longitudinal-mode selection, and output-power scaling through distinct photonic strategies.

An adjacent development replaced cavity redesign with system-level power combination. Two $2.5\,\mathrm{cm}$ Er:TFLN waveguide amplifiers were coherently combined on chip using 2×2 MMIs and an integrated EO Mach–Zehnder phase shifter. With $6\,\mathrm{mW}$ seed at $1530\,\mathrm{nm}$ and $15.8\,\mathrm{mW}$ total $1480\,\mathrm{nm}$ pump, the coherently combined output reached $12.9\,\mathrm{mW}$, with on-chip net gain $3\,\mathrm{dB}$, power conversion efficiency $38\%$, and EO performance $V_\pi=8.3\,\mathrm{V}$ for a $3\,\mathrm{mm}$ phase shifter, or $V_\pi L=2.49\,\mathrm{V\cdot cm}$ [2308.08740]. This suggests that, on Er:TFLN, “laser” performance can be improved not only by resonator engineering but also by combining amplification and EO phase control on the same gain-bearing chip.

## 4. Wavelength management and electro-optic programmability

The broad erbium pump and emission bands create a need for pump–signal routing on the same chip. A dedicated example is the angled multimode interferometer wavelength-division multiplexer fabricated on $0.5\,\mathrm{mol}\%$ Er:TFLN. Designed for resonant pumping around $1480\,\mathrm{nm}$ and emission in the $1530\text{–}1560\,\mathrm{nm}$ band, it reported on-chip insertion losses below $0.7\,\mathrm{dB}$ in both wavelength ranges, minimum signal-port insertion loss $\approx0.68\,\mathrm{dB}$ at $\sim1540\,\mathrm{nm}$, pump-port insertion loss $\approx0.52\,\mathrm{dB}$ at $1480\,\mathrm{nm}$, $3\,\mathrm{dB}$ bandwidth $>20\,\mathrm{nm}$ in the C band, and pump-port suppression more than $18\,\mathrm{dB}$ lower than the signal port across the signal band [2401.05970]. Because the same device also visualized multimode interference through erbium upconversion fluorescence, it directly linked the active dopant to passive wavelength-routing diagnostics [2401.05970].

Electro-optic programmability is the other defining control layer. In laser-oriented papers the relevant first-order resonance relation is expressed as $\Delta\lambda/\lambda\approx\Delta n/n$, with $\Delta n$ governed by the Pockels effect [2108.09740]. In practice, this has already been used for EO tuning of an integrated Er:LNOI microring laser at $0.33\,\mathrm{pm/V}$ [2103.08197]. The most advanced demonstration, however, appears in the cryogenic $^{167}\mathrm{Er}^{3+}$ quantum-memory platform. There, gold electrodes on TFLN waveguides yielded measured tuning efficiency $1.11\,\mathrm{GHz/V}$, static preparation of four independent spectral channels with total tuning up to $2.69\,\mathrm{GHz}$, and dynamic routing between two channels by square-wave voltages of $\pm0.8\,\mathrm{V}$ at rates up to $20\,\mathrm{MHz}$, while maintaining inter-channel crosstalk below $10^{-4}$ [2605.14777]. The same work used the standard LN relations
$$
\Delta n = -\frac{1}{2}n^3r_{33}E,\qquad \Delta f \approx -f_0\left(\frac{\Delta n_{\mathrm{eff}}}{n_{\mathrm{eff}}}\right),
$$
to interpret cavity-frequency control [2605.14777].

The broader significance is that erbium provides the optical transition, but LN determines how that transition is routed, tuned, and multiplexed. This is a structural distinction from erbium-doped passive hosts: on Er:TFLN, wavelength selection can be integrated at the same lithographic level as gain and resonant storage. A plausible implication is that future “active-passive” partitions on the chip may become less rigid, because pump combiners, filters, modulators, and erbium gain sections already coexist in the reported device literature.

## 5. Cryogenic spectroscopy, implantation, and quantum memory

At cryogenic temperatures, Er:TFLN becomes a platform for telecom-band light–matter interfaces rather than only gain media. Post-implantation studies first established that implanted erbium could coexist with high-Q TFLN photonics. After $350\,\mathrm{keV}$ implantation and post-implant annealing at $550^\circ\mathrm{C}$, microring resonators recovered average loaded $Q\approx5\times10^5$, with observed loaded $Q$ near $10^6$ in one device. A $4.7\,\mathrm{cm}$ waveguide exhibited a room-temperature photoluminescence peak near $1532\,\mathrm{nm}$ with FWHM $\approx250\,\mathrm{GHz}$, the cryogenic excited-state lifetime was $T_1=3.2\,\mathrm{ms}$ at $1.7\,\mathrm{K}$, and cavity-enhanced decay in a ring gave an ensemble-averaged Purcell factor of $\sim3.8$ with $T_{\mathrm{ring}}=618\,\mu\mathrm{s}$ and average coupling rate $g_{\mathrm{av}}\approx2\pi\times597\,\mathrm{kHz}$ [1912.07584].

Focused-ion-beam implantation on x-cut TFLN then added deterministic spatial control. Implanted $^{170}$Er regions retained Stark-split ${}^4f$–${}^4f$ transitions consistent with bulk Er:LiNbO$_3$, with representative low-temperature lines including $Y_1\rightarrow Z_1$ at $1530.9\,\mathrm{nm}$, $Y_1\rightarrow Z_2$ at $1545.4\,\mathrm{nm}$, $Y_1\rightarrow Z_3$ at $1561.7\,\mathrm{nm}$, and $Y_1\rightarrow Z_5$ at $1573.9\,\mathrm{nm}$ after rapid thermal annealing at $850^\circ\mathrm{C}$ for $60\,\mathrm{min}$ in flowing O$_2$ [2511.04442]. Cooling from $295\,\mathrm{K}$ to $\approx100\,\mathrm{K}$ increased both integrated PL intensity and lifetime, but below $\approx50\,\mathrm{K}$ the integrated intensity fell by $\approx50\%$ between $50\,\mathrm{K}$ and $5\,\mathrm{K}$ while lifetime decreased by $\approx10\%$, behavior attributed to suppression of the host pyroelectric response and the resulting internal-field changes [2511.04442]. In that interpretation,
$$
\Delta P \approx p_3 \Delta T,\qquad E \approx \frac{P}{\epsilon_0\epsilon_r},
$$
and the erbium Stark shift is written as
$$
\Delta E = -\vec{\mu}\cdot\vec{E} - \frac{1}{2}\alpha E^2.
$$
This establishes that cryogenic Er:TFLN behavior cannot be reduced to rare-earth spectroscopy alone; thin-film boundary conditions and host ferroelectricity measurably enter the optical response [2511.04442].

The most complete quantum-memory realization used isotopically purified $^{167}\mathrm{Er}^{3+}$ in an X-cut TFLN racetrack microring at $260\,\mathrm{mK}$. The native telecom transition is centered at $1532.13\,\mathrm{nm}$ with inhomogeneous broadening $202\,\mathrm{GHz}$, and long-lived hyperfine shelving states support atomic-frequency-comb preparation with persistent single-component comb lifetime $T_{\mathrm{AFC}}=277.6\pm52.6\,\mathrm{s}$ and spectral preparation efficiency $\eta_{\mathrm{spectral}}\approx0.95$ [2605.14777]. For $100\,\mathrm{ns}$ storage with tooth spacing $\Delta=10\,\mathrm{MHz}$ and Gaussian input pulses of $15\,\mathrm{ns}$, the on-chip storage efficiency was $23.3\pm0.5\%$, while temporal multiplexing reached $16.5\pm0.3\%$ collective efficiency for nine $5\,\mathrm{ns}$ pulses within $100\,\mathrm{ns}$ and $6.0\pm0.2\%$ for up to eighteen modes in $200\,\mathrm{ns}$ [2605.14777]. The echo timing followed
$$
t_{\mathrm{echo}}=\frac{1}{\Delta},
$$
and the cavity-enhanced efficiency model was reported as
$$
\eta_{\mathrm{AFC}}=\left[A(F,\eta_{\mathrm{spectral}})\times\left(\frac{\kappa_{\mathrm{ext}}}{\kappa_{\mathrm{total}}}\right)\times\left(\frac{4C'}{(1+C')^2}\right)\right]^2\times \eta_d.
$$
The same device stored and retrieved time–energy-entangled telecom photons with on-chip storage efficiency $8.6\pm0.5\%$, cross-correlation $g^{(2)}_{si}(0)=4.54\pm0.30$, Franson visibilities $V_1=51.17\pm1.19\%$ and $V_2=51.30\pm1.21\%$, and an entanglement witness $W=-0.1037\pm0.0092$, violating the separable bound by more than $11\sigma$ [2605.14777]. Within the Er:TFLN literature, this moves the material system from classical active photonics into integrated quantum networking.

## 6. Constraints, trade-offs, and development directions

Several limitations recur across the reported devices. Coupling loss remains a dominant systems bottleneck: early amplifiers used lensed-fiber coupling losses of $\sim7.3\,\mathrm{dB}$ per facet at $1550\,\mathrm{nm}$ and $\sim10\,\mathrm{dB}$ per facet at $980\,\mathrm{nm}$ [2101.00783], while the electrically pumped hybrid microring laser reported $\sim10\,\mathrm{dB}$ butt-coupling loss from the pump diode into the Er:TFLN waveguide [2209.01533]. Later work showed that integrating spot-size-converter edge couplers with $<2\,\mathrm{dB}$ UHNA7-to-chip coupling and roughly $3\,\mathrm{dB}$ lumped SMF-to-chip loss per facet is sufficient to convert high internal gain into $>18\,\mathrm{dB}$ external net gain and output powers above $13\,\mathrm{dBm}$ [2508.11941]. This suggests that external usefulness is now limited less by the existence of gain than by the efficiency with which that gain is connected to the rest of the optical system.

Internal trade-offs are equally explicit. In the coupled microdisk–microring single-mode laser, increasing the disk–ring gap raised threshold from $1.31\,\mathrm{mW}$ to $2.72\,\mathrm{mW}$ and $3.18\,\mathrm{mW}$ while reducing slope efficiency, because weaker coupling left more energy in microdisk-only WGMs [2108.09740]. In the large-mode-area amplifier, the $1\,\mu\mathrm{m}$-wide guide saturated at $\approx18\,\mathrm{mW}$ output and then collapsed as pump exceeded $\approx200\,\mathrm{mW}$, whereas the $9\,\mu\mathrm{m}$ guide continued to scale to $\approx113\,\mathrm{mW}$, making mode area itself a control parameter for ESA- and CUC-limited power handling [2405.16583]. High gain can also create new failure modes: parasitic lasing was observed when on-chip gain exceeded $\approx28\,\mathrm{dB}$ at $1531.5\,\mathrm{nm}$ and $\approx21\,\mathrm{dB}$ at $1550\,\mathrm{nm}$ in a high-power EDWA, and angled facets plus index-matched edge couplers were proposed to suppress these back-reflection-induced cavities [2405.16583].

Material-specific nonlinearities and host effects remain important. Pump-induced photorefractive phase shifts reduced coherent-beam-combination extinction ratio from $\sim15\,\mathrm{dB}$ without pump to $\sim13\,\mathrm{dB}$ with pump in an EO-controlled two-amplifier combiner [2308.08740]. Enhanced photorefractive response in an Er:TFLN photonic-crystal nanobeam produced a measured cavity blueshift of $\approx-1.38\,\mathrm{nm/mW}$ under $974\,\mathrm{nm}$ pumping [2312.15601]. At cryogenic temperature, the x-cut implanted platform exhibited a low-temperature PL anomaly linked to pyroelectric-field suppression rather than to a conventional rare-earth relaxation channel [2511.04442]. Linewidth and long-term stability are still incompletely documented in many laser reports; for example, the WGSML paper did not report linewidth, and the Sagnac-loop FP laser linewidth was explicitly OSA-limited [2108.09740] [2212.12698].

The reported development directions are therefore highly specific. For amplifiers, they include lower-loss fabrication, optimized Er concentration and profiles, high-efficiency edge couplers, and cladding- or LMA-based pump engineering [2111.05571] [2405.16583] [2508.11941]. For lasers, they include refined coupling geometry, deterministic wavelength control, EO electrode integration, and cavity designs that combine low mode volume with better output extraction [2103.08197] [2309.10512] [2312.15601]. For quantum devices, the proposed path is toward intrinsic cavity quality factors $>10^7$, deeper AFC preparation, stronger magnetic-field and anneal optimization, and on-demand spin-wave storage, with the explicit estimate that reducing cavity loss could move on-chip memory efficiency beyond $70\%$ [2605.14777]. Across all of these directions, the central point remains unchanged: erbium turns TFLN from a passive electro-optic substrate into an active telecom-band photonic material, but the practical performance envelope is set by how successfully gain, low loss, cavity engineering, and EO control are co-optimized on the same thin-film lithium-niobate platform.

Source: https://www.emergentmind.com/topics/erbium-doped-thin-film-lithium-niobate