5.5 Thin film: lithium niobate on insulator (LNOI)

L1 Phenomenon

Helium or hydrogen ions shot into a LN wafer stop at a depth set by their energy, about 0.7 µm for 195 keV helium. The wafer is flipped and bonded to oxidized silicon. Gentle heating makes the implanted gas collect in a buried plane until the crystal splits there, leaving a sub-micrometre single-crystal film on the oxide. Polishing removes the damaged top, and annealing repairs the lattice. This is the same "Smart Cut" family of processes that makes silicon-on-insulator wafers.

Ion slicing: implant depth and thermal stepsR_p 0.68 µm

(a) projected range vs energy, LiNbO₃

Ion

He⁺

Energy E195 keV

E 195 keVR_p 0.68 µmreported

(b) temperature at each step

StepImplant

True depth of the sourced point: He⁺ density peak 0.68 µm below the implanted face at 195 keV reported, 0.14% of a 500 µm donor; no straggle is sourced, so no width is drawn. He⁺ 195 keV, 4 × 10¹⁶ cm⁻², holder 100 °C reported. Open circles at split: 150 °C, 2 h (H in LN) P; 190 °C, 10 h (LiTaO₃) P.

680 nm films were reported by ion slicing in 2004 without the energy in the abstract P(abs); H⁺ 100 keV, 7 × 10¹⁶ cm⁻² is reported for LiTaO₃ only P. Tilt 7° against channelling P.

Ion

Implant energy195 keV

Process stepImplant

Stopping depth: 0.68 µm
Step temperature: 100 °C

He⁺ at 195 keV stops at a 0.68 µm density peak reported. Range data exist only for the reported 195 keV He⁺ point; there is no verified range table for He⁺ or H⁺ in LiNbO₃. Not to scale.

Process viewer stepsopen in the Process viewer

#	STEP	ROUTE
07	Prepare the handle wafer (Si with thermal SiO2)	ridge
08	He+ implantation into the LN donor	ridge
09	Clean, plasma-activate, flip, and bond donor to handle at room temperature	ridge
10	Strengthening anneal, then split at the implanted plane	ridge
11	Edge removal and CMP of the split surface to target thickness	ridge
12	Recovery anneal of the film (wafer supplier step)	ridge

                                                
LLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLL
LLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLL
LLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLL
LLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLL
LLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLL
LLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLL
LLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLL
LLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLL
                                                
                                                
                                                
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
LLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLL
LLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLL
LLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLL
LLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLL
oooooooooooooooooooooooooooooooooooooooooooooooo
oooooooooooooooooooooooooooooooooooooooooooooooo
oooooooooooooooooooooooooooooooooooooooooooooooo
SSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSS
SSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSS
SSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSS
SSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSS

Step progress t0.85

Step

Step progress0.85

10/23Strengthening anneal, then split at the implanted plane (ridge route)

L2 Model and data

Lineage

Year	Step	Source; status
1995	Smart Cut for SOI (H implantation plus bonding)	P(title)
1998	Crystal ion slicing of LN: deep implantation makes a buried sacrificial layer in poled c-cut wafers; 9 µm films bonded to Si and GaAs; same dielectric and pyroelectric constants and T_c as bulk	P(abs)
2004	680 nm films by ion slicing and wafer bonding on SiO₂ over electrodes; r₃₃ = 31 pm/V at 633 nm	P(abs)
2007	BCB-bonded z-cut films; microring R = 100 µm; 0.14 GHz/V; EO response about 50% of bulk because BCB limits annealing to 320 °C (full recovery needs about 800 °C)	P
2009	LNOI photonic wires	P(title)
2012	LNOI wafers of 3-inch diameter by ion slicing and bonding	P(abs)
2021	Ion-cut LNOI review (implantation, bonding, material)	P(abs)
as of 2026-10-24	300-900 nm LNOI on 3, 4, 6, 8 inch (NanoLN); LN6 at 150 mm (Soitec POI)	V, V

Ion implantation

Quantity	Value	Source; status
He+ energy, fluence, temperature (z-cut CLN)	195 keV; 4 × 10¹⁶ cm⁻²; holder at 100 °C	reported
He+ density peak at 195 keV	0.68 µm	reported
LNOI convention	He+ above 200 keV	P
H+ for LTOI (analog)	100 keV, 7.0 × 10¹⁶ cm⁻², 4-inch x-cut LiTaO₃; beam current ten times higher than for He	P
Tilt to avoid channelling (H, LN, 4-inch)	7° from normal	P
Range vs energy curve for He and H in LN	not tabulated in any source read	N/A

Bonding, splitting, recovery

Step	Value	Source; status
Pre-bond clean and activation	RCA clean; N₂ plasma activation; room-temperature direct bonding (EVG 301)	P
Bond-strengthening anneal	below 200 °C	P
Split	about 230 °C	P
Split, after RT bond (H-implanted LN on Pt/SiO₂/LN)	150 °C in N₂ for 2 h	P
Split (LiTaO₃ analog)	190 °C for 10 h	P
Handle used for TFLN photonics	Si with 4.7 µm thermal SiO₂ (500 or 525 µm); quartz 500 µm; LN substrate with 2 µm SiO₂	P, P, P
Post-split surface	edge removal and CMP remove the rough, implantation-damaged layer and thin to target (LiTaO₃: to 600 nm)	P; P
Device-level recovery anneal	O₂, atmospheric pressure, 520 °C, 2 h; material-limited loss of ion-sliced LN reduced from about 1.5 dB/m to about 0.2 dB/m	P

Thickness metrology (wafer-scale)

Quantity	Value	Source; status
Edge exclusion for film-thickness mapping, 4-inch	8 mm (rough rim from wafer production)	P
Etch-depth standard deviation, 300 nm etch, 4-inch	5.9 nm (3.2 nm excluding a further 6 mm)	P
LTOI thickness map (analog)	600 to 630 nm colour scale across the wafer (axis range of Fig. 1c)	P

Alternative routes

Route	What is reported	Source; status
Bonding plus grinding and CMP, no implantation	LN bulk bonded on 2 µm silica, thinned by CMP to 4 µm; microrings with Q about 10⁸	P
Thick films (5-50 µm)	offered commercially	V
BCB adhesive bonding	large-area (> 1.5 cm²) films; limits anneal to 320 °C	P
Epitaxy (1974)	epitaxial LN waveguides for beam scanning	P(title)
Sputter deposition	sputter-deposited TFLN RGB multiplexer (638, 520, 473 nm)	P(abs)
Chemical beam epitaxy	c-axis LN on C-plane sapphire at 400 °C, for SAW	P(abs)
Atomic layer deposition	ferroelectric LiNbO₃ by ALD	P(title)
Sol-gel nanoimprint	nanoimprinted LN for nonlinear metalenses	P(title)

CMP after the split: thickness and uniformity100 mm, 8 mm edge exclusion

thickness deviation map (template field, σ = 5.9 nm); x: excluded rim

          xxxxxxxxxxx          
      xxxxxx%%%%%%%xxxxxx      
    xxxx#######****###%xxxx    
   xxx******++++++++**##%xxx   
  xx+++============+++*##%@xx  
 xx===-------==---===+**#%@@xx 
xxx=----:::--------==+**#%@@xxx
xxx---:::::::::----=++*##%@@xxx
xxx=---:::::::----==++*#%%@@xxx
xxx===------------==++*#%@@@xxx
 xx++========-----==++*#%@@@xx 
  xx**+++++=========+**#%@@xx  
   xxx#****+++++++++**#%@xxx   
    xxxx###*********##%xxxx    
      xxxxxx#######xxxxxx      
          xxxxxxxxxxx

Shading " .:-=+*#%@" from −12 to +12 nm about the mean. The paper's map is not reproduced; this field carries only its reported standard deviation. template

Split depth (He⁺ 195 keV)	680 nm	reported
Target film	600 nm	reported
Removed	40 of 80 nm	derived
σ, 8 mm excluded	5.9 nm	P
σ, further 6 mm excluded	3.2 nm	P
Preston rate K_p P v	120.0 nm/min	template
Time to here	0.3 min	template
Film colour (D65)	#989ea9	derived

Film thickness h 640 nm

K_p (template), Pa⁻¹1.0×10⁻¹³

Pressure P (template) 20 kPa

Velocity v (template)1.00 m/s

dz/dt = K_p P v (Preston 1927). No Preston coefficient for LN is sourced: K_p, P and v are display parameters, so the rate and time are template. Film colour: normal-incidence reflectance of air / LN / 4.7 µm SiO₂ / Si under D65 with the Zelmon 1997 indices, CIE 1931 colour matching, sRGB.

Film thickness640 nm

Colour at the mean: #989ea9
Removed so far: 40 nm
Spread (reported σ): 5.9 nm
σ, further 6 mm excluded: 3.2 nm

K_p, pressure and velocity are not sourced for LN, so no removal time is given here template.

Colour of air / LN / 4.7 µm SiO₂ / Si under D65 at normal incidence, from the Zelmon 1997 indices; the 4.7 µm oxide adds dense fringes that pull the colour toward grey, so the shift from 680 to 600 nm is subtle derived. Thickness field template.

L3 Details and open points

Why the film must be annealed: implantation damage raises absorption P; post-slicing recovery of r₃₃ needs high temperature P. Annealing also risks Li₂O out-diffusion, one hypothesis for higher coercive fields in thin films P.
Thermal mismatch: LN expands 14.1 ppm/K along a and 4.1 ppm/K along c P-comp; in x-cut films both axes lie in the plane, so the film strain on a Si handle is anisotropic. A sourced Si handle CTE is not held: N/A.
Which face ends up on top: the implanted surface becomes the buried interface; the split surface is on top and is polished. For x-cut, flipping about an in-plane axis reverses the on-screen direction of +Z or +Y (Process viewer, step 9).
680 nm (Rabiei) and 0.68 µm (Guarino peak) coincide numerically; whether Rabiei used 195 keV is not stated in the abstract.

← 5.4 Waveguides in bulk wafers 5.6 Wafer sizes and suppliers →