Establishing a Failure Analysis Lab
- Written by Tom Paquette
Have you've ever wondered what it might take to start your own semiconductor Failure Analysis (F/A) lab, whether as an internal lab or as an independent lab? There are many considerations, most importantly, the main purpose or goal of the lab.
For example, the need for Integrated Circuit (IC) cross sectioning has dramatically different requirements than if you needed Focused Ion Beam (FIB) edit capability. Although they seem at the extreme ends of complexity, both have their challenges.
In order to establish the capability to cross section ICs, the equipment needs are often understated. Obviously there are the lapping wheels, which can easily cost as much as $20,000 new, or older used models may be purchased off the internet for at a more modest price (about $6,000). The complexity of the technology is a major factor in this decision. A second wheel should also be seriously considered to better facilitate the lapping process by allowing multiple grit papers/solutions to be available at the same time.
But before one can begin cross sectioning the part, it often must be removed from its package (a depot is a procedure to remove the plastic packaging material, leaving the silicon die unscathed). The most straight-forward method to depot a plastic encapsulated IC is through wet chemical etching, using a small quantity of nitric acid or a mixture of nitric and sulfuric acids. Naturally a chemical hood is required, plus a minimal number of acids and solvents, vented cabinets for the chemicals, glassware, personal protective equipment, and a book of Safety Data Sheets for each of the chemicals.
Local agencies will require inspections of the wet chemical area and a plan to neutralize the acid waste and to dispose of the solvent waste must also be developed.
In order to examine the cross section, an optical microscope is often used during the course of the cross section and a scanning electron microscope (SEM) for the final inspection. An optical microscope with digital image capture is recommended; this will typically cost anywhere from $6,000 (used) to $45,000 for a new state-of-the-art system.
For many mainstream integrated circuits technologies, the high-magnification capabilities of a Field-Emitter SEM (FE-SEM) are required. Typically a new FE-SEM is priced in the $100's of thousands of dollars, alternatively refurbished equipment may be available (plan on about $45,000 for an older system). In order to image the features in the cross section, it is often necessary to "stain" them. Staining is an etching process where the cross section is partially immersed in different acids (typically) specifically tailored to etch the materials in the cross section in order to deliniate features of interest. Staining is critical for SEM visual inspections and is nearly always wafer-fab process dependent. While there are many industrially-shared staining recipes, many are often considered company trade secrets and kept confidential.
In addition to the FE-SEM, a coater system is required. Coaters are systems that deposit a thin layer of a conductive material on the surface of the cross section to help ground the sample while in the SEM. In most instances, a gold coater is sufficient (used ones can sometimes be found for about $4000), although for inspections greater than 100kx an Iridium coater is mandated (as the gold coating will begin to be imaged in the SEM at high magnifications), at a cost of about $75,000.
Other expenses that need consideration are the lapping and chemical supplies (plan a budget of several thousand dollars for initial supplies), and maintenance contracts (which run $10,000 to $50,000/year for a SEM or FE-SEM, respectively). In all, the equipment to support IC cross sectioning and inspection can run in excess of $400,000. Lab space and facilitating the equipment are other considerations.
In comparison, an IC submitted to an outside failure analysis lab for cross sectioning might cost in the neighborhood of $1,000 each (including SEM inspection). If a return on investment of 3 years is the goal, an internal need for about 800-1,000 cross section jobs in that timeframe would be needed to justify the cost (including internal manpower) of bringing the work inside.
Now let's look at the FIB, a new FIB starts at about $600,000 for a system to edit IC designs; the maintenance contract will add roughly another 10% of the initial cost each year (after the warranty period) In this case, the samples may be bare silicon or packaged (necessitating the chemical lab requirements already outlined). Before installation, the area planned to house the FIB will have to be audited for electromagnetic interference, vibration, and acoustic interference by the vendor. Also, unless your company already has the facilities, the FIB will require a source of chilled water and should also have an Uninterruptable Power Supply (in case power is lost, which could cause harm to the equipment). The proper UPS system can easily cost $10,000.
The sophisticated lab equipment outlined so far is only half of the recipe for a semiconductor failure analysis lab, the proper staff to run the equipment is critical to the success of the facility. No universities in the USA offer degrees in failure analysis, so individuals use the knowledge acquired while earning degrees in physics, chemistry, Material Science, and electrical (or other) engineering programs to enter this field and earn their F/A "degree" on the job. As such, skilled failure analysts are rare and often in high demand, therefore one should plan on at least a year to find the right person(s). If a recent graduate is hired, the industry rule of thumb is that it will take 2-3 years of training before they are strong failure analysts (depending upon the complexity of the process and job).
In summary, in order to establish basic failure analysis capabilities for an in-house or independent lab, the costs, facilities, maintenance, supplies, inspections, and personnel selections must be taken into account. For basic IC cross sectioning capability, the budget should be planned from about $100,000 (used equipment) to $400,000. For FIB editing, the budget should be closer to $750,000. In all practicality, one should plan about a year to have the lab equipment and staff in place and ready to produce.