We do not share your results with anyone who is not listed on your request form. Please, if you wish us to share your results with your colleagues or even your boss, we will happily do so if they are listed on the request form granting us permission to do so.   We need a completed analysis request form before profiling your samples. Our request form is your communication to us. Including as much information as possible will help us to properly perform SRA on your wafers. A detailed description of region of interest and if applicable, a photograph is always helpful. An appropriate bevel angle is needed to provide optimum resolution in your main area of interest. Two profiles may be needed, depending on the size of the pattern. We may also need to produce two bevels as well. We use beveling blocks with different angles depending on your structures needs.

Above is an illustration of a shallow and deep profile Above is an illustration of a shallow and deep profile. Are you interested in the shallow portion of the structure or the deep portion or maybe both? We need to know if there is any specific region of the profile of greater importance than the rest?

Of course, we have the capabilities of measuring smaller than a millimeter square, but it is more difficult. Four-point probe sheet resistance measurements are generally measured on non-patterned wafers or on patterns which are a minimum of 3mm x 3mm. It is always helpful to provide backup samples. For example, when cutting out the sample, sometimes they break right in the middle of the area you are trying to profile, resulting in a smaller pattern and forcing us to use a steeper angle and therefore jeopardizing the resolution. Beveling beyond the pattern is also a concern especially if you have only one pattern to work with.   We use various beveling blocks with different degrees of arc depending on your structures needs. The blocks are mounted on a stainless steel jig and placed on a conditioned glass plate. Our experience has been that long waiting periods, such as overnight or even an hour or two tend to produce noisier data, due to native oxide grown on the beveled sample. This is especially true with p-type dopant. The minimum requirement of 20m is due to our probes spacing. Our probes can only be placed so close together without shorting out. At Solecon our average probe spacing is 17-18um apart. Although we do have the ability to have narrower spaced probe tips but they tend to break faster. The reason why we have a 100um length requirement is we cannot run our probe steps smaller than 2um increments without landing on damage created by previous measurements, and the smaller the pattern length the resolution is jeopardized. The size of the probe contact area is 2um wide x 2-5um high.  

Spreading resistance test patterns offer advantages such as a complete characterization of the wafer fab process, optimum resolution and a clear straight-forward identification of all possible dopant structures.

  Our probes tips are made of tungsten carbide. The bevel edge line represents the transition between the top surface and the beveled surface. We always start the scan on the top surface and run down into the silicon. We deliberately probe a few points on the original surface so that we are sure we do not miss any data at the start of the bevel. When we reduce the profile, the data points on the top surface are discarded.

  After probing your samples, an additional scan is performed with one probe tip heated slightly. This additional scan is as close to the original scan as we can. We estimate that during contact, the silicon underneath the hot probe is heated to 1 to 2 degrees Celsius. A small voltage (known as the “Seebeck Voltage”) is produced across the probe tips. The polarity sensed at the hot probe is equal to the polarity of the minority carriers in the silicon. This determination has been found to be effective for all but very high resistivity values – presumably due to electrons haveing greater carrier mobility than holes.

  Our optical profilometer is like an atomic force microscope, capable of measuring roughness, step heights, and other critical dimensions. This profilometer will also give us the opportunity to look at the topography of the wafer. It has played a big part in our study of bevel edge rounding as well as our study of our grinding media (the glass plates) we use for beveling your samples.

Only the very front portion of the probe arms (extending from top-left to the center of the photograph) can be seen. Even so, they dwarf the sample which is mounted on to the angle block. At the top-right is a 40X microscope objective providing an idea of the size of everything. The probe tip on the rear probe is the one that gets heated during the hot-probe operation. One can barely see the outline of a wire on the rear probe arm. This is part of the hot-probe wiring.

Given a cylinder of uniform resistivity Given a cylinder of uniform resistivity. When good ohmic contact is made to both ends of a cylinder, the resistance measured is equal to the resistivity multiplied by the length and divided by the cross-sectional area. However, when the size of the contacts is reduced, the measured resistance increases dramatically. When reduced far enough, the length of the cylinder no longer has significant impact on the measured resistance. The measured resistance approaches the resistivity divided by twice the radius “a” of the contact area. This is due to the “current crowding” or “spreading resistance” in the region immediately (first one or two microns) under the contacts. Since this small region (referred to as the “sampling volume”) controls the measured resistance, one can determine the resistivity in this region once the contact radius has been determined. Because the length is of no consequence for small contacts, both contacts can be brought to the same side of the cylinder and be brought close to each other with little change in the measured resistance. In practice, contacts made using specially shaped probe tips. We use tungsten carbide probe tips but other hard, low resistivity material might also be used.

As the thickness of the layer decreases (we referred to this as “length” in the previous slide), the measured resistance begins to increase. When the thickness “t” becomes less than a few microns, the measured resistance becomes a function of the sheet resistance rho/t, the probe spacing “s”, and the electrical contact radius “a”. (In this case, the thickness of the entire layer is being “sampled”.)

Conductivity and resistivity need to be defined Conductivity and resistivity need to be defined. In practice, there is usually a great disparity between the electron carrier concentration “n” and the hole concentration “p”. Except in cases of very light doping, the smaller of the two (referred to as “the minority carrier”) can be set to zero thus making the equation a function of the majority carriers only. This simplification cannot, however, be used for very high resistivity material.

Here is how we determine the electrical contact radius, “a” Here is how we determine the electrical contact radius, “a”. Sixty-four silicon standards of well documented resistivity are used. Half of the standards are n-type – half are p-type. Half are <111>, half are <100>. The thickness of the standards is orders of magnitude thicker than “a” thus a “bulk” measurement (rho /a) is produced. “a” varies depending on the weight applied and shape of the probe tips so that each set of probe tips must be calibrated and checked often.

Calibration data for three probes is shown Calibration data for three probes is shown. As expected, heavily loaded probes have a larger contact radius “a” and therefore have lower measured resistance for a given resistivity. As expected, <111> silicon has a smaller contact radius than <100> due to the high density of atoms on the plane. Typically, p-type silicon has a lower measured resistance than n-type for a given resistivity. The reason for this is still a matter of controversy.

Three samples of different sheet resistances having a thin conductive layer at the surface were measured on the same probe at varying probe spacings. When plotted on a semi-log graph, the points extrapolate back to 2 microns. Therefore, the contact radius must be 2 microns for this set of probe conditions.

To get minimum probe spacing, material must be removed (ground away) from the side facing the other probe tip.

The next few slides will illustrate the depth profiling of a not intentionally doped epitaxial layer over an n-type substrate. First, the sample is angle lapped (beveled). The sample is then placed under the probes such that the bevel is in a horizontal plane. The probing is started on the original surface. The probes are VERY gently lowered to the silicon. Five millivolts are applied across the probe tips. The resistance is measured. The probes are then lifted and the sample indexes one step increment to the left. The probes are again gently lowered. A plot of measured resistance is produced and this plot is commonly referred to as the “raw data”.

Using a proprietary data reduction system, the raw data is translated into a resistivity-depth profile. The data reduction system must have provisions to deal with the bevel angle, the step increment, the calibration data, and sample volume correction. A considerable amount of mathematics is required.

The resistivity data is converted to carrier concentration using the carrier mobility values derived from Thurber, Mattis, Liu, and Filliben, National Bureau of Standards Special Publication 400-64, The Relationship Between Resistivity and Dopant Density for Phosphorus-and Boron-Doped Silicon (May 1981), Table 10, Page 34 and Table 14, Page 40. If the sample has lower than expected carrier mobility due to crystal damage or other reasons, the calculated carrier concentration will be low.

Usually, one can assume that dopant concentration and carrier concentration are the same BUT there are at least two important exceptions: 1) For structures commonly know as “wells” (moderately doped and extending relatively deep) the electrical junction is shallower than the metallurgical junction and 2) Structures with a concentration “dip” at the surface. Before going onto the next slide, we need to appreciate that the dopant concentration is calculated from the carrier concentration which in turn, is calculated from the resistivity profile. These additional calculations increase the uncertainty somewhat.

We believe spreading resistance senses the electrical junction at the top of the depletion region rather than the metallurgical junctions where the concentration of donor atoms equals the concentration of acceptor atoms. Note the deeper junction in the dopant profile.

Part of the difference in the tail region is due to the electrical junction being shallower than the metallurgical junction. This study was done by Dr. James Ehrstein of the National Bureau of Standards (NBS) – now known as N.I.S.T.

The profile has the expected LSS shape The profile has the expected LSS shape. Its depth, however, is extremely impressive! Arsenic ions in the 5+ charge state were accelerated by a 2.2 megavolt terminal voltage to produce this impressive structure shown in figure 7. The implantation was done at the Lawrence Berkeley Lab* Peter Byme et. al. "Megavolt Arsenic implantation into Silicon, 1982 International Conference on Metallurgical Coating and Process Technology.

The most likely scenario for this profile is that the antimony in the buried layer region was contaminated with about 1% boron. During the buried layer drive-in, the faster diffusing boron out-runs the antimony producing a big bulge below the N+ buried layer. Also note, there is a depression in the concentration at the epi-buried layer transition. This too could have been caused by the boron contamination. During the epitaxial deposition, a smaller amount of thermal energy was expended than in the buried layer drive-in so, less out diffusion of boron is expected. Further thermal processing of the wafer may cause a p-layer to form thus defeating the intended purpose of the buried layer.

Arsenic Autodoping:   Arsenic and boron have sufficient vapor pressures at epitaxial deposition temperatures to produce significant autodoping during the start of epi growth. Figure 6 shows two spreading resistance profiles, measured on the same bevel of a sample, immediately after an atmospheric pressure, SiCl4 epi deposition. The first profile, taken through the buried layer pattern, indicates an epi thickness of about 3.5 µm with appreciable intrusion of the buried layer dopants up into the epi layer. This profile has an often seen, non-gaussian, steep fall off in the tail region. This is often the signature or an arsenic buried layer. Referring to the second profile, taken through the epi substrate, we see a ghost buried layer. The epi-thickness as indicated by the peak concentration is essentially the same 3.5 µm while the p-n junction is about 0.8 µm deeper and the peak concentration is roughly five times as great as the intentional epi concentration. This ghost buried layer would tend to cause the V/I on the epi test wafer to read deceptively low if done on a multi-wafer reactor.

The figure at the right (zero degree tilt) shows a channeling “tail” which more than doubles the junction depth as compared with the figure at the left (10 degree tilt). The shape of the profile in figure 8b is similar to that obtained by depositing and diffusing a slow and fast diffusing species (e.g., antimony and phosphorus) concurrently. Frankly, we were surprised.

These profiles were taken after a P+ (boron) deposition and drive into an n-type epi layer about 12 microns thick. The profile at the left easily connects with the p-substrate producing a robust “isolation”. In the profile on the right, the isolation is very marginal. ( The hole concentration in the transition region is less than in the substrate.)

A light p diffusion is applied to the surface of a p-type (Czochralski) substrate having a concentration of 2E14 atoms per cubic centimeter. The wafer is then subjected to a number of hours at 400C. When oxygen-rich wafer (Czochralski grown ingot) is oxidized, oxygen at the silicon surface (surprisingly) is removed forming a “de-nuded zone”. Then, during the 400C treatment, some of the oxygen in the rest of the wafer forms “thermal donors” counter-doping the boron. Note the majority carrier concentration (holes in this case) has been drastically reduced. In this example, the resistivity is uneven. There are extreme cases, however, where the carrier type is completely converted to n-type with the exception of the de-nuded zone if present.

Prior to anneal (dopant activation and crystal repair), spreading resistance usually will not sense an implant dose less than 1E13 ions per cubic centimeter. At higher doses, however, there tends to be enough damage to the crystal that the increase in resistivity (due to reduced carrier mobility) can be sensed. In this example, the measured resistance (“raw data”) has increased somewhat at the surface. The profile on the right shows a calculation of the carrier mobility. An intermediate profile of resistivity depth was obtained. Then, using the bulk concentration as a constant, the carrier mobility was calculated from the resistivity profile.

There are certain structures where it is impossible to obtain optimum resolution for each layer with one bevel. A shallow source-drain with a deep well into epi is one such structure. To get the source-drain, a shallow bevel is needed. For the rest of the structure, a steeper bevel is needed. The center profile focuses on the well and the remaining epi. The profile on the right shows the up-diffusion of the p+substrate into the epi. Very little of the epi remains unaffected by the p+ up-diffusion.