GB2190212A - Semiconductor electro-optic device with controlled birefringence - Google Patents

Abstract

The device comprises a layer of semiconductor material (2), for example a silicon wafer, in which an array of transverse parallel slits (1) is formed. The device therefore exhibits birefringency to radiation having a wavelength more than twice the spacing of the slits and the birefringency effect is governed by controlling the free carrier density within the semiconductor. This may be effected by applying an electric potential at (6, 4) between respectively the body of the semiconductor layer and the surfaces (3) of the slit walls, which are doped (7) to a polarity opposite to that of the body of the layer. When such a device is assembled between two prisms a switchable polarizing beam-splitter is formed (Fig. 3). Other applications are switchable quarter- and half-wave plates, and shutters. <IMAGE>

Description

SPECIFICATION Structures having birefringement properties This invention relates to structures incorporating arrays of slits, and in particular to optical devices having such structures in which the width and spacing of the slits is small compared with the wavelength of the applied electromagnetic radiation.

It is known that when the spacing of the slits in such structures is less than half the wavelength of the incident radiation, then such structures appear uniform to the radiation and also exhibit birefringency. This is because, under these conditions, the refractive indices np and n, parallel and transverse to the plane of the slits in a material having a bulk refractive index n can be shown to be respectively (q+n2(1-q))i and (q+n-2(l-q))-, where q is the ratio of slit width to repeat distance.

Various devices using this effect may be produced, the properties depending on the wavelength of the radiation and material used, for example. Thus, if slits are formed in silicon, the values found for np and nt are 2.85 and 1.60 respectively for q=1/3. Slits having a width of 2 ,um may be usefully etched in silicon wafers, so that devices for use at 8 to 14 iim wavelengths may be fabricated.

Such devices are nevertheless passive, in that in any single device n and q are fixed, so that the birefringence effect cannot be varied.

However, if the refractive index of the medium can be varied, so can the birefringence.

It has been found that the refractive index of semiconductor materials, such as silicon, is to some degree dependent on the free carrier density of the material, and this invention accordingly consists of a device comprising a layer of semiconductor material having formed in it an array of parallel slits, and means for controlling the free carrier density in the material forming the walls between the slits.

It is possible to use silicon wafers as the semiconductor medium, and although the maximum phase differential between orthogonal components of radiation incident on the face of the wafer is proportional to the distance travelled through the depth of the slits, a phase change of about 10" may realistically be aimed for by capacitative depletion of free carriers using slits 40 ,um deep, 2 ijm wide formed at 4,um pitch (q=2) which are suitable, as mentioned above, for 8 to 14 ,'m wavelength radiation. The slits may be produced, for example, by anisotropic etching techniques.Such a device between two suitably orientated polarizers will perform as a shutter with no moving parts having, on the application of a low voltage, a switching speed in the order of IOns.

Even with a 10 shift a shutter action which may be of use can be achieved, but its on/off action would be poor or its transmission small. If an enhanced effect is required, then a multiple passage through the silicon may be arranged, although this is limited by the transmission coefficient of the medium. This can be achieved, for example, by making use of multiple internal reflections.

The variation in carrier density within the semiconductor may be achieved by current injection, but although inferior in effect, capacitative methods require a much lower power consumption, and the essential structure of such a device formed in a silicon wafer is illustrated in Fig. 1, which is a schematic partsection transverse to the wafer and in the plane of a slit.

A large number of parallel slits 1, each 2 ,um wide, are formed in a (110) silicon wafer 2, p-doped to 5-15 ohm cm, at about 4 ,um pitch so that the walls between them are also about 2 Hm wide. A n-doped layer 3 is diffused into the surface of the walls and floor of the slits, each such diffusion being connected via a terminal 4 bonded to a deep surface n±diffusion 5 along a line beyond the ends of the slits to a variable bias voltage Vb.

The other terminal 6 is connected to the wafer and hence the underlying slit walls. As Vb is varied, the carrier density within the slit walls is depleted or enriched, and the refractive index of the material, and hence the birefringence effect of the device, varies correspondingly. It may be advantageous to increase the dopant concentration in the silicon by initially forming an increased p-doped layer 7 around the slit walls and floor.

The fabrication of such a structure can be divided into several stages as follows: 1. The n±diffusions 5 are formed prior to the slits being etched. Firstiy, 3500A of silicon dioxide is grown on the front and back of the wafer and resist is applied. The resist is removed by photolithography from the areas where the diffusions are to be located and the oxide layer dissolved in 15:1 buffered HF. The remaining resist is then removed. Arsenic, as the dopant, is implanted from the front surface at 100 KeV to 5x 1015 cm--2 before the remaining oxide is removed in buffered HF.

The implant is annealed at 1050"C in argon for 30 minutes.

2. Next, the slits 1 are etched, after a similar cycle of growing an oxide, photolithography and etching the oxide has defined the slit pattern, using an anisotropic etch such as ethylene diamine-pyrocatechol-water (potassium hydroxide-water-alcohol mixtures could be equally effective except for the problems of potassium contamination). At etch of 1 41 hours at 80"C will produce slits about 35 Am deep and provided that the slit masks have been carefully aligned with the crystal planes the slit walls will not be thinned too much by sideways etching.

3. The surface concentration of p-type dopant 7 can be boosted by a boron diffusion, for example for 90 mins at 950"C. The oxide and any glassy layer on the surface are removed in buffered hydrofluoric acid, after which a 1 00A oxide layer is grown and a drive-in diffusion of boron and arsenic is carried out, typically for 16 hours at 1100 C.

This anneal results in the boron being driven in to a depth of 1.4 ,am and the arsenic to 1.0 ,um.

4. The surface n-diffusion 3 is formed, after protection of the front by LTO (which does not penetrate the slits), by diffusing in phosphorus as POCI5 for 5 minutes at 900"C.

At this stage, the back face is also given an optional arsenic and phosphorus implantation, both to 5x 1015 cm-2 at 180 KeV for gettering purposes.

5. A 400A thick silicon dioxide layer is then grown on the front surface of the wafer, and removed by photolithography and etching from those areas where electrical contact is to be made.

6. Aluminium contacts are formed by depositing a 1 ,um thick layer on the front surface at 45" to the normal, to avoid significant deposition within the slits. After photolithography (best done using a negative resist process), the unwanted areas are removed by etching in a phosphoric, nitric and acetic acid mix. The remainder is lastly fired in at 450"C for 1 hour in forming gas.

One problem in carrying out the above fabrication process is that once the slits have been etched, resist deposited prior to subsequent photolithography stages tends to drain into the slits and an even surface covering is difficult to obtain. This can be overcome either by using an inorganic resist or by depositing a sufficiently thick layer of material such as silicon dioxide to bridge the slits. If the latter method is used a 3 ,um thick layer of silicon dioxide can be deposited by chemical vapour deposition, the upper 1 ,um of which is phosphorus-doped so that, when heated, it flows into the grooves formed over the underlying slits to form an even top surface. The oxide layer so formed may then be etched in the same way as the oxide layers are produced, as described earlier.

It is also possible to arrange that no photolithography has to be done after the etching of the slits. An alternative method of fabricating a similar structure thus consists of the following stages, in which reference is made to Figs. 2a to 2d, which are schematic sections of a wafer in the plane of a slit: 1. After a pre-oxide clean, a p-type, 5-15 ohm cm (110) silicon wafer 11 has a 2000A thick layer of silicon dioxide grown on its front and back surfaces by steam oxidation at 900"C. The bonding area 12 for the eventual 'slit' electrode (contact 13 see Fig. 2d) is delineated by photolithography, lightly etched to produce alignment marks, and implanted with phosphorus at 100 keV to 2x 1015 cm-2.

After dissolution of the oxide and amorphous layer in buffered HF, the phosphorus is driven in by annealing in argon at 1150"C.

2. A further 2000A thick layer of silicon dioxide 14 is grown. The back surface is protected by resist, a pattern corresponding to the slits to be etched is produced in resist on the front surface by photolithography, and the oxide layer removed in buffered HF where exposed. The structure is now that illustrated in Fig. 2a.

3. In turn, a 1000A nitride layer 15 and a 1 500A low temperature oxide layer 16 are then deposited. Voids in the oxide layer are reduced by baking in steam at about 105000.

4. A resist pattern separating the contact areas (one of which includes the entire slit area) is produced by photolithography, and the oxide in the contact areas is removed in buffered HF before the remaining resist is removed. After the nitride thus exposed is removed in hot phosphoric acid, the structure is that illustrated in Fig. 2b, in which the oxide layer 16 and the nitride layer 15 overlie parts of the areas of oxide 14 deposited on the silicon substrate 11.

5. The slits are formed by etching to a depth of 25 Am in EDA at 80"C, and thoroughly removing afterwards all traces of the etch. After boron has been diffused in for 10 minutes at 815"C, the structure of Fig. 2c results: a p-doped layer 17 covers the walls and floor of the slits 18.

6. The exposed oxide and amorphous layers are removed in buffered HF, and a further 1 00A oxide layer is grown by steam oxidation at 900"C to provide protection for the next stage.

7. The boron is driven in by a bake at 1050"C for 121 hours in argon and the oxide exposed around the nitride pattern is removed in buffered HF.

8. Phosphorus is diffused in for 5 minutes at 800"C to produce n-doped areas 19 and 20 (see Fig. 2d) overlying the p-doped surface of the slits and in the areas at which contacts 13 and 21 are to bonded to the wafer. Contact 21 is made to the p-substrate (breaking through the thin n-layer 20 in the bonding process) and contact 13 to the n+ layer.

One application of devices fabricated by these means is in beam splitters, an example of whose operation is illustrated schematically in Fig. 3. The device illustrated consists of two triangular prisms 31 and 32 stuck together by a layer 33 having a lower refractive index. According to the particular refractive indices, incident light, such as that shown in the figure, will either be totally internally reflected within the first prism or pass through the layer 33 and the second prism 32. If the layer 33 is birefringent, then one polarization may be transmitted and the other reflected The dependency of internal reflection or transmission on the value of the refractive index of the layer 33 relative to that of the prism 31 gives rise to the possibility of using the slit structures as the material used for the layer 33.

Thus, such a device consists of two machined or micro-etched prisms, and a silicon wafer constructed in accordance with the present invention forming the middle layer. Care naturally has to be taken to ensure that the proper internal reflections are involved, but by this means the application of a relatively low voltage can be made to control very rapidly whether the incident light is reflected or transmitted, and the structure therefore becomes a very efficient optoelectronic switch.

Another application of the invention is in the production of half and quarter-wave plates.

The total phase lag of one polarisation relative to the other depends on the carrier density and thus on the applied voltage. The effect of changing the voltage will typically be to change the structure from a 52 to a 5.52 plate, depending on the wavelength, depth of slit, etc.

This is acceptable if the device is to act as an electronically-switched wave plate at just one wavelength, but a broader bandwidth can be achieved by mounting two identical devices with their slits orthogonal to one another on top of each other. When no voltage is applied, the phase advance of a given polarisation in one set of slits is equal to the retardation in the other set, so no overall birefringence effect is produced. When a voltage is applied to one but not the other, the structure becomes birefringent, and so an electronicallycontrolled birefringent structure results. The effect on the polarisation of applying a voltage is similar over a substantial range of wavelengths.

By combining an electronically controlled birefringent structure (ideally switching between a zero and a half wave piate) with two polarisers, a switch can be produced. For example, if the polarisers are at 450 to the slits, and are orthogonal to one another, radiation will not be transmitted when the structure is acting as a zero wave plate but will be transmitted when it acts as a half wave plate.

Further applications of this invention, which will be readily apparent to those skilled in the art, include electronically adjustable quarter and half wave plates and rapidly acting shutters to protect light-sensitive equipment.

Claims (8)

1. A structure for the transmission of electromagnetic radiation comprising a layer of semiconductor material having in it an array of parallel slits formed in planes normal to the plane of the layer, and means for controlling the free carrier density in the material forming the walls between the slits and hence for controlling the birefringence effect exhibited by the device to radiation whose wavelength is more than twice the spacing of the slits.

2. A structure according to Claim 1 in which the layer of semiconductor material is a silicon wafer.

3. A structure according to either preceding claim in which the free carrier density is controlled by capacitative depletion.

4. A structure according to Claim 3 in which the surfaces of the walls of the slits are doped to the opposite polarity to that of the layer of semiconductor material and the capacitative depletion is achieved by applying an electric potential between the said surfaces and the body of the said layer.

5. A device comprising at least identical structures according to any preceding claim in which the free carrier density in each structure can be controlled independently.

6. A device comprising two prisms separated by a structure according to any of claims 1 to 4 whose transmission of electromagnetic radiation is governed by the free carrier density in the structure.

7. A structure substantially as hereinbefore described with reference to Fig. 1.

8. A method of fabricating a structure for the transmission of electromagnetic radiation substantially as hereinbefore described with reference either to Fig. 1 or to Figs. 2a to 2d.