When you send a request for quotation to a TFLN wafer supplier, five parameters determine whether the wafer you receive will actually work in your photonic integrated circuit: crystal cut, film thickness, buried oxide (BOX) thickness, wafer diameter, and whether the lithium niobate substrate is congruent-melt or MgO-doped. Underspecify any of these and you are not requesting a wafer — you are requesting a surprise. This article explains what each parameter means, how to choose it, and what the choices cost you in lead time and availability.

Crystal Cut: X-cut vs Z-cut

The crystal cut of the lithium niobate film determines how the ferroelectric c-axis is oriented relative to the wafer surface, and that orientation determines the electrode geometry you need to access the dominant electro-optic coefficient.

In X-cut TFLN, the c-axis lies in the plane of the wafer. Electrodes placed laterally on either side of the waveguide ridge apply an electric field that is largely parallel to the c-axis, exploiting the r33 coefficient (approximately 30 pm/V), which is the strongest electro-optic tensor element in lithium niobate. Because the electrodes do not need to be placed directly above the waveguide, the optical mode avoids the metal absorption that would otherwise add insertion loss. X-cut is the dominant configuration for electro-optic modulators targeting coherent optical communications and microwave-photonic applications.

In Z-cut TFLN, the c-axis is perpendicular to the wafer surface. This geometry is used in some nonlinear-optical and acousto-optic devices where the vertical field orientation is structurally convenient. Fabrication is more involved because the electrode must cross over the waveguide, requiring careful design of the buffer layer to avoid excessive optical absorption from metal proximity.

For most integrated photonics procurement today — modulators, switches, IQ modulators for coherent transceivers — X-cut is the right starting point unless your device architecture specifically requires Z-cut geometry. Most foundry process design kits (PDKs) for TFLN target X-cut substrates.

Film Thickness: 300–600 nm Range

The thickness of the lithium niobate film is one of the most consequential parameters in your device design. It sets the optical mode size, which in turn determines bend loss, coupling loss to fibre, and the overlap between the optical field and the RF field applied by the electrodes.

Thinner films (300–400 nm) confine the optical mode more tightly. This benefits integration density — you can achieve sub-millimetre bend radii — and enables high confinement in nonlinear-photonic applications where tight mode overlap with a periodically poled structure is desirable. The trade-off is that thinner films are more sensitive to waveguide sidewall roughness (since the evanescent field is comparatively larger relative to total mode power), and coupling into standard single-mode optical fibre becomes more difficult because the mode mismatch is larger.

Thicker films (500–600 nm) relax the lithography and etch requirements, improve fibre-chip coupling via intermediate-geometry spot-size converters, and typically result in lower propagation loss for a given etch quality. The trade-off is that tight bends require more real estate and the RF-optical overlap factor — critical for modulator half-wave voltage Vπ — changes. Most foundries with commercially mature TFLN processes converge on 300–600 nm as the practical range, with 400 nm and 600 nm being the most commonly catalogued stock thicknesses.

Buried Oxide (BOX) Thickness

Beneath the lithium niobate film sits a layer of thermally grown or deposited silicon dioxide, which forms the lower optical cladding and electrically isolates the active layer from the silicon handle wafer. This layer must be thick enough that the evanescent tail of the guided optical mode does not leak into the high-index silicon substrate — a loss mechanism that worsens as film thickness decreases and mode confinement loosens.

Standard BOX thicknesses for TFLN range from approximately 2 µm to 4.7 µm. For a 600 nm LN film, a 2–3 µm BOX is generally sufficient for telecom-wavelength operation. For thinner films (300–400 nm) where the mode extends further into the cladding, 3–4.7 µm BOX provides a safer margin against substrate leakage loss. If you are designing devices at shorter wavelengths (near-infrared or visible), the leakage problem intensifies and you may need to specify the thicker end of the BOX range.

BOX thickness also affects bonding stress and wafer bow. Thicker oxide layers introduce more stress mismatch between the lithium niobate film, the oxide, and the silicon handle, which can contribute to wafer bow — a parameter that becomes critical in lithography steps that require tight focus across the full wafer diameter. When requesting wafers, ask your supplier for bow and warp specifications alongside the film and BOX parameters.

Wafer Diameter: 4-inch vs 6-inch

The wafer diameter determines which foundry processes are accessible to you and significantly affects per-die cost at volume. Four-inch (100 mm) TFLN is the most widely available format today, supported by a broad range of suppliers. Six-inch (150 mm) wafers are available but from a narrower supplier base, and specification control — particularly film thickness uniformity — can be harder to verify across the full 150 mm area. Eight-inch TFLN is in early-stage development at select foundries as of 2026.

For prototyping and low-volume research, 4-inch wafers offer the widest supplier choice and the most catalogue availability. If you are planning a volume transition to a merchant foundry running CMOS-compatible tooling on 6-inch or larger, plan your qualification and supply-chain sourcing process to accommodate a longer lead time for the larger format.

MgO Doping

Lithium niobate is susceptible to the photorefractive effect — a reversible but practically disruptive phenomenon in which absorbed optical power causes charge redistribution that modifies the local refractive index, destabilising device behaviour. MgO doping at roughly 5 mol% raises the photorefractive damage threshold by several orders of magnitude, making MgO:LN substrates the required choice for high-power applications, devices operating at wavelengths below approximately 1 µm, and second-harmonic or sum-frequency generation where visible light is generated within the waveguide.

For standard telecom-band electro-optic modulators operating at 1310 nm or 1550 nm with modest optical power levels (below a few hundred milliwatts in the waveguide), congruent-melt lithium niobate without MgO doping is usually adequate and offers a slightly wider supplier base. Specify MgO-doped TFLN if your device will handle high optical powers, will operate at visible or near-visible wavelengths, or if the application involves extended operation where even slow photorefractive drift is unacceptable.

Putting It Together: A Minimal Specification

A complete TFLN wafer specification for a standard telecom-band modulator development project should read, at minimum: X-cut, 600 nm LN film ± 10 nm uniformity (full-wafer), 3 µm SiO2 BOX, 500 µm silicon handle, 4-inch diameter, congruent-melt substrate, film non-uniformity <1% (1σ), wafer bow <30 µm. Add MgO doping and expand BOX thickness to 4.7 µm if operating below 1 µm wavelength or at elevated power.

Supplier datasheets rarely volunteer all of these numbers in a single document. A deep-tech sourcing engagement — one that systematically compares suppliers against a specification matrix rather than accepting the first datasheet as a bid — is the structural solution to a market where the quoted headline number rarely represents the worst-case across the full wafer area.

TFLN晶圆规格的核心参数包括：晶向（X-cut vs Z-cut）、薄膜厚度（300–600 nm）、掩埋氧化层厚度（2–4.7 µm）、晶圆直径（4英寸或6英寸），以及铌酸锂基底是否掺MgO。以下逐一解析每个参数的选择逻辑与代价。

晶向：X-cut与Z-cut。 X-cut TFLN的c轴位于晶圆平面内，侧向电极即可利用最强的电光系数r33（约30 pm/V），无需金属直接覆盖波导芯。Z-cut的c轴垂直于晶圆表面，电极需跨越波导，工艺更复杂，但适合部分非线性光学应用。大多数电光调制器应用首选X-cut。

薄膜厚度：300–600 nm。 较薄（300–400 nm）有利于紧密集成，但对侧壁粗糙度更敏感，与光纤耦合也更难；较厚（500–600 nm）降低了工艺难度和传播损耗，但弯曲半径需要更大。目录库存多为400 nm和600 nm两种规格。

掩埋氧化层（BOX）：2–4.7 µm。 BOX隔离光学模式与高折射率硅衬底，防止底部漏光。600 nm薄膜配2–3 µm BOX通常足够；300–400 nm薄膜或短波长应用需要3–4.7 µm。同时务必索取晶圆翘曲数据。

晶圆直径：4英寸 vs 6英寸。 4英寸TFLN供应商最多，6英寸供应商较少，均匀性控制难度更高。原型阶段建议4英寸；批量生产转6英寸需提前规划更长的合格化周期。

MgO掺杂。 约5 mol% MgO可将光折变损伤阈值提高数个数量级，适合高功率、短波长（<1 µm）或对光折变漂移零容忍的应用。标准电信波段（1310/1550 nm）调制器通常无需掺MgO，且可获得更广的供应商选择。

一份完整的最简规格示例：X-cut、600 nm薄膜 ±10 nm（全片均匀性）、3 µm SiO₂掩埋氧化层、500 µm硅衬底、4英寸直径、同成分熔融基底、晶圆翘曲 <30 µm。供应商数据表很少在单份文件中给出所有参数，深度技术采购的核心价值在于系统比较而非接受第一份数据表作为报价依据。

TFLN晶圓規格的核心參數包括：晶向（X-cut vs Z-cut）、薄膜厚度（300–600 nm）、埋氧化層厚度（2–4.7 µm）、晶圓直徑（4英寸或6英寸），以及鈮酸鋰基底是否摻MgO。以下逐一解析每個參數的選擇邏輯與代價。

X-cut TFLN的c軸位於晶圓平面內，側向電極即可利用最強的電光係數r33（約30 pm/V），無需金屬直接覆蓋波導芯。Z-cut的c軸垂直於晶圓表面，電極需跨越波導，工藝更複雜，但適合部分非線性光學應用。薄膜厚度方面，300–400 nm有利於緊密整合但對側壁粗糙度更敏感，500–600 nm降低工藝難度但彎曲半徑需更大。BOX需足夠厚以防底部漏光：600 nm薄膜配2–3 µm BOX通常足夠，300–400 nm薄膜需3–4.7 µm。MgO摻雜可大幅提高光折變損傷閾值，適合高功率或短波長應用。標準電信波段調制器通常無需摻MgO。深度技術採購的核心價值在於系統比較供應商規格，而非接受第一份數據表。