2) When do I need sample throughput and should not consider a SEM conversion?
RAITH150, RAITH50 and especially SEM-based solutions are not suited for high-throughput mask making work or high-throughput production oriented type of work in industrial scale. They are targeted for small scale production, for device optimization and, for example, research like quantum devices etc. Please consult Raith to find out if your application is still within the possibilities of Raith equipment.
3) When should I consider a complete system like RAITH150 and not a SEM conversion kit?
The answer has to be short, because we recommend to discuss your intended application with us. If you prefer a SEM retrofit and would like to buy the SEM along with the e-beam attachment, you may have to deal with two or more companies and buy SEM from company A, motorcontroller from company B, and e-beam system from RAITH. Although the final order can be commissioned at the SEM vendor, this does not mean, that they are able to guarantee the complete system integration. Details of integration or testing still will require you to support yourself, even though RAITH will give best support with great commitment. Furthermore, it will be difficult to give high guarantees, e.g. for parameters like stitching and overlay, which are mainly influenced by SEM electron optical performance and not RAITH system performance. It is impossible to have all system combinations tested that you may like to decide on! Additionally, if one has to build "one off" solutions, e.g. for stages, redesign may become necessary, which can have an impact on other project schedules. With RAITH150 and RAITH50 we have now some years of experience with electron optical systems. Furthermore, the mechanical components like the high-precision laser stage are designed and tested even for large travel ranges. Thus we are able to tell you about the guaranteed specification before you place an order. Last but not least, the support will be coordinated by RAITH and there is no "shifting of responsibility" between parties.
4) What is the resolution limit of the lithography system?
Electron beam lithography is capable of producing sub 5 nm structures using even the most economic ELPHY Quantum type solution exposing contamination resists. This resolution is more than magnitude below the current semiconductor manufacturing standards. However, the limiting factor for practical and working devices is pattern transfer like lift off or etching of certain substrates. For many years PMMA has been used for high-resolution applications but in practice has reached limitations in the sub 20 nm range. Using multilayer techniques and special development conditions, researcheres now achieved sub 10 nm resolution using PMMA. New types of resist like Calixarene are under evaluation and have qualified for sub 10 nm resolution as well.
5) What are the main differences between a "self-made" lithography attachment and a professional ELPHY system?
If you consider building your own solution out of components like DAC board and software, the advantage can be the overall investment, if you do not have to pay the people who do this work. As long as it is well documented, even following generations of users may be able to use the system satisfactorily. We recommend to check this alternative carefully in regard to participation in continuous commercial product development and user support. RAITH's latest generation software can update systems that have already been in use for many jears! Please consider also that you keep upgrading possbilities for state-of-the-art systems or even laser stage technology.
6) Can I use modern Field Emission (FE) SEM for e-beam lithography?
The answer is yes! However, COLD FE SEM technology has been on the market more than 15 years, and those systems went through quite a strong evolution. COLD FE SEM are offering outstanding beam brightness and resolution, but due to the characteristics of the electron generation the electron emission is not too stable. As a result, the probe current and therefore the exposure dose can differ by up to 30 %. Only special precautions will allow you to use cold FE SEM in a relatively stable region for a few hours. Furthermore, cold FE SEM are restricted in regard to maximum available probe current. Basically one can say that in several installations cold FE SEM has proven to be an excellent tool for high-resolution exposures, if you understand to operate them in a stable region. They are not to be recommended for "long term" exposures like prototype masks or larger area exposures.
Generally, the advantage of FE SEM for lithography is that the excellent resolution is very helpful to adjust beam focus and stigmatism. The intensity of the beam is much stronger than in conventional SEM. During the last few years THERMALLY ASSISTED FE SEM has proven to be the most interesting alternative for e-beam lithography. Due to the thermally assisted electron generation, the beam stability and the maximum probe current issue have been solved. Most relevant manufactures also offer an integrated electrostatic beamblanking solution.
7) Does low voltage lithography makes sense?
Today′s R&D is also focussing its efforts on understanding pattern transfer techniques using low-voltage exposure. The main motivations can be found in the uncertainty of the industry about reaching the sub 100 nm design rules for semiconductor production. If particle beams will be used, the main problem will be to overcome the principle limitations of throughput. There are several approaches on parallel electron column arrays and parallel probe techniques being evaluated. But they all have in common that high energies cannot be applied to miniaturized electron optics. Today′s FE columns offer outstanding sub 100 nm capabilities, which now allows a wide field of new academic and technology-oriented R&D to contribute to the success of new lithography techniques.
8) What is the advantage of having a high writing speed?
Throughput is not the main issue of SEM based e-beam lithography, but a high writing speed is also important for other reasons as is shown below: A modern SEM column with thermal field emission allows to focus an e-beam with 100 pA beam current into a spot of much less than 10 nm in diameter. Even when using a very insensitive resist with 100 μC/cm2 the dwelling time in each written pixel must be less than 1 μs corresponding to 1 MHz writing speed. If such writing speed is not available, compromises have to be made by reducing the beam current, which leads to a lesser image quality (or alignment capability) etc. Many resists with higher sensitivity require much more caution, if a proper exposure is possible at all. For very special applications ELPHY allows also to control the dwelling time of each pixel down to the ps range by using a special, very fast blanker and special pulse electronics. But this cannot be a general solution due to the resulting long total exposure times (including blanking times) and its own disadvantages of drift etc. Therefore Raith lithography systems offer up to 10 MHz writing speed in true area exposure mode and not just applying a dedicated exposure mode using fast beam blankers.
9) Shall I prefer high voltage on my lithography system?
If you ask around in the community, some people will say, if I expose at 50 kV or 100 kV I can achieve much better pattern sharpness and I can achieve much higher resolution. This is true to the extent that in the past the resolution of electron optical systems increased with the energy. In addition, it is true that higher energy minimizes the forward scattering of electrons in thick resist. For some application, which require working in thick resists (Microsystems) this has been proven to be advantageous. Disadvantages of high-energy beams are beam damage, and due to insensitive resists overheating of resists can be a problem. Nevertheless, electron optical technology went through quite an evolution, so today you can have outstanding electron resolution even at very low voltages.
10) Why is a two-dimensional laser interferometer stage superior to any stage with precise measurements in two axes?
High positioning accuracy is wanted for the e-beam in relation to the sample, but this cannot be measured directly except by mark recognition using the e-beam itself. Extremely precise measurements in two single axes i.e. using linear encoders do not help much, because the extrapolation to the beam position will be influenced strongly by inaccuracies of the stage movement (yaw, pitch, roll), which cannot be avoided. The best approach is a two-dimensional laser interferometer stage, where the two measuring laser beams reflected at a fixed L-shaped mirror virtually meet at the spot where the electron beam strikes the sample. With the so-called "ABBE condition" fulfilled, the influences of yaw, pitch and roll will be eliminated in the first order. Therefore the question is not how precisely X and Y can be measured, but how precisely the measured coordinates represent the real position of the sample under the e-beam. If you can rely on your stage's accuracy, this will allow to set up e.g. Mix&Match exposures for gate level writing fully automated. With a laser stage you won′t need to do mark registrations for each single writing field, which helps to minimize the writing time and totally avoid registration overhead. Short writing times may make the need for drift compensation obsolete.
Stitching means that patterns are written that exceed one of the fundamental limitation of basic lithography systems: the size of writing field. Depending on the application, maximum useful and low-distortion writing fields will not exceed 500 μm to 1 mm. If you want to expose over larger areas (e.g. prototype masks) with precise matching on the writing field borders, your writing field and stage translation must be calibrated to each other. This is only possible with laser-interferometer metrology!
12) Should I consider second-hand SEM for lithography? Which SEM should I buy?
Second-hand SEM - even old ones - have proven their capability to generate patterns even in the sub 50 nm range. They are a valuable alternative, if budgets are restricted. Many groups made their initial state-of-the-art R&D by converting old instruments. Due to the initial results applications for modern FE SEM have been successfully approved. ELPHY systems can be easily moved to the new instrument, and therefore to attach ELPHY to an old SEM is a non-risk investment.
13) Why do I need a hierarchical GDSII editor - can I not simply import patterns from commercial CAD systems?
Indeed import from other commercially available design software or even ASCII data is possible by using standard import facilities (DXF, CIF, GDSII) for users who prefer to work with AutoCAD or are using advanced IC design software. However, doing so has often been rated not very user-friendly because one has to assign exposure dose values for patterns either as import default or, e.g. in AutoCAD or similar programs by using layers or colors for dose assignment. In some cases one has to use other separate software packages for postprocessing, fracturing or proximity effect correction. This means sometimes painful data import and export and sometimes incompatibility of versions. Finally, when tha data has been converted into the individual e-beam machine format, the user may realize that the pattern still needs shape or dose modifications - so the loop has to start again...
With the RAITH hierarchical GDSII editor the user may completely forget about pattern data compilations using 3rd party packages. He can assign exposure doses, modify pattern shapes and can start the exposure without bothering about any low-level e-beam machine format - it is very straightforward. The capability to use a hierarchical format greatly helps to reduce the amount of pattern data, if repetitive or complex pattern, are designed. Reading GDSII is also very easy if you use the graphical user interface of ELPHY to prepare patterns e.g. for a Mix&Match exposure.
14) What are typical values for pattern overlay accuracy?
The final overlay accuracy depends on the chosen writing field size and detector arrangement. As a rule, overlay accuracy of converted SEM should typically expected to be better than 1/1000 of the size of the writing field, e.g. 500 nm in a 500 μm writing field. It has been shown that in a 100 μm writing field the overlay accuracy inside the writing field is typically better or around 100 nm when markers are used. With some experience on the overall system this value can be achieved in even larger fields.
15) Why is it good to have a FAST and DEGLITCHED 300 KHz (at least!) hardware?
Unlike any other PC board type pattern generator solution, RAITH has designed its own proprietary hardware for its lithography systems, which on the one hand allowes us to offer higher writing speed than standard 16 bit DAC boards and, on the other hand we took care of the dynamic behavior of the main deflection DACs. DEGLITCHED DAC means that RAITH made some efforts to reduce output voltage spikes occurring e.g. when most significant bits of DACs are switched.
16) What is the advantage of "hardware alignment" and "calibration" capability over software means only?
Basically, alignment is used to align subsequent exposure steps after treating the sample with different technological steps (development, evaporation, annealing, etching). Unfortunately, it is not sufficient to rely only on mechanical means of the lithography systems stage to achieve the requested overlay accuracy. Basically, one has to shift the electron beam relative to the sample, which can be done with sufficient resolution to achieve good overlay results. There are two solutions: either is to calculate a truly displaced or distorted pattern, or the solution RAITH prefers is to use a set of "multiplying" DACs being responsible for a real analog writing field shift, rotation and scaling. In older PCs the hardware alignment basically was much faster than a simple software-based solution, where some data crunching was necessary. This difference decreases with modern fast PCs, but to use multiplying DACs still has advantages, e.g. for field calibration. Hardware alignment also allows to use the maximum DAC range for exposure. The 16 bit multiplying DACs in combination with the normal 16 bit DACs provide a pattern definition accuracy of 32 bit. Usually this seems not to be important, because the real placement accuracy of single features within a writing field is certainly not better than 14 bit. Applications with high duty grating like DFB lasers, where the extremely high definition accuracy is needed, should be written in a fairly large field size, if no stitching capability provided by a laser stage is available. Assuming you need to choose a 500 μm field size, the minimum pattern placement step will be 500 μm/16 bit = 7.63 nm. DFB applications require being able to perform a fine tuning of the average pitch in the sub nm range, in order to filter out the desired laser wavelength. With multiplying DAC the remaining pattern placement step can be scaled with 16 bit resolution again, which approximates 0.01%.
17) Why is it important to have high writing speed resolution?
If you do low-voltage exposures (resist sensitivity increased! ) or if you really need to apply a high-fidelity proximity correction, you have to consider that, depending on the hardware architecture, the minimum dwell can be quite low, e.g. 0.5 μs. But if the writing system finally only accepts to change dose values in the range of e.g. 0.1 μs, one can easily see that the next exposure available value would be e.g. 20 % of 0.5 μs, which is far too much to allow a good proximity correction. One can certainly consider to change the probe current, but practically this increases the complexity of usage of the the e-beam system. RAITH ELPHY Plus hardware implemented means that it can accept less than 10 ns dwelling time values, which is about 3 % of the maximum writing speed of ELPHY Plus. Within RAITH 150 this resolution is in the range of 2 ns!