Maxwell11.2 March 2008
Moore with less silicon
How the silicon substrate slowly comes alive
There has been much attention for the tremendous developments
integrated circuit technology has went through. The focus is always
on the size and speed of the devices. In past years, however, the
substrate itself also underwent remarkable developments. This has
resulted in silicon-on-glass technology and flexible circuits. The
fu-ture holds even more, with stretchable circuits and even LivingChips
on the horizon. The following article presents an overview.
Author: Prof.dr.ir. Ronald Dekker, Philips Research / TU Delft
In 1965 Gordon Moore made a remar-kable prediction. He predicted that every two years or so, the number of transistors on a chip would double. Forty years later his prediction still holds. A remarkable feat, especially since the maximum num-ber of transistors on a chip was only 50 when Moore formulated his law, while it is exceeding 2 billion for the latest de-signs! [1] To make all this happen, extre-mely advanced and expensive production tools have been developed. Probably best known are the wafer steppers; the photo-lithographic tools that are used to define features as small as 45 nm.
Less known is that, besides the produc-tion tools, also the development of silicon substrates has gone through a tremen-dous development. Whereas in the early days circuits were processed on 1 inch wafers, today silicon wafers with a diame-ter of 300 mm are standard, while 450 mm is ready in research. Compared to wafer steppers, the development of these silicon substrates receives only little pu-blicity. Nevertheless, they belong to the most perfect and purest objects made by man. A blank silicon wafer is essentially a single crystal with a purity and perfection far better than that of diamond, at a price of less than 100 euro.
From a silicon technologist perspective, silicon is nearly an ideal material to work with. It is strong, cheap, easily forms a stable isolating oxide, it is a good thermal conductor and can be etched in various ways. From an electrical perspective, it has a good mobility for carriers, a band gap that is high enough to prevent leakage of transistors at normal temperatures and a resistance that can be accurately con-trolled over seven orders of magnitude, just by incorporating tiny amounts of im-purities in its crystal.
What is often not realised is that in most integrated circuits everything of interest is usually confined to the top one or two mi-cron of silicon. Comparing the thickness of a silicon wafer to the height of a table, this layer has the thickness of a sheet of paper! The remainder of the silicon sub-strate just serves to keep the whole thing together during fabrication. Fortunately, in most cases the material properties of the silicon substrate are such that they don’t interfere with the functioning of the circuit. Silicon, for instance, is a good thermal conductor, which is helpful to re-move the heat from the circuit layer to the package for high speed processors. However, for a lab on chip, where for in-stance a small compartment has to be
heated for a certain reaction, the leakage of heat through the thermally conductive substrate can be a real problem. Also for RF circuits the conductive silicon substra-te underneath the circuits introduces los-ses and an undesired coupling of different circuit parts.
Subtrate Transfer
Substrate Transfer combines the ease and advantages of standard silicon processing, with a complete freedom of substrate choice. Basically, fully processed silicon wafers are glued, top down, to a new sub-strate with properties more suitable for the application. Next the original silicon substrate is removed by a combination of grinding and wet-etching down to an etch stop layer directly underneath the circuit. In this way the thin layer containing the circuits is transferred to the new sub-strate. For many applications the circuits are transferred to glass substrates. Glass is a good thermal and electrical isolator, and its transparency allows for ultraviolet curing of the adhesive. The strength
Figure 1: Glass wafer containing record per-formance varactor diodes made in DIMES.
Q
10
Maxwell11.2 March 2008of the technology is that it is a so called “post-processing” technology which does not interfere with the normal IC proces-sing.
The technology was developed at Philips Research and later transferred to NXP Hamburg, where it was subsequently in-dustrialised. In Delft, Lis Nanver from ECTM-DIMES immediately recognised the advantages of Substrate Transfer for RF circuits. Together with the staff from DIMES, she explored the fascinating
pos-sibility of double sided device processing that Substrate Transfer offers. Recently, Leo de Vreede and Lis Nanver demon-strated varactor diodes on glass with a world-record performance, again showing the advantages of a substrate free of para-sitics. [2]
Flexible circuits
During the development work at NXP in Hamburg, an accidental discovery was made. In order to improve bondability of the circuits, in an experiment a 10 µm
thick layer of polyimide (Kapton) was ap-plied to the wafer before it was glued to the glass substrate. However, a fortunate mistake was made! The wrong procedure was used to ensure an optimal adhesion between the polyimide and the adhesive. During the subsequent wet etch that was used to remove the silicon substrate, the circuit layer attached to the 10 µm thick film of polyimide spontaneously delami-nated. To everybody’s surprise the tran-sistors and circuits on the foil still func-tioned perfectly.
At Philips Research the procedure was modified to control the moment of dela-mination. The reliability of these flexible circuits is obviously a concern. Extensive experiments have shown that as long as the circuit layer is held under a slight compressive stress, these circuits can be bent in either direction to a radius of less than 1 mm without breaking of the cir-cuit layer. A 10 µm thick RFID chip for chip-in-paper applications was designed and fabricated (Fig. 3). This chip, which constitutes the worlds thinnest IC, re-mains fully functional, even while bended to very small radii.
The flexibility of these ICs makes them ideally suitable for biomedical applicati-ons. Due to their flexibility, these chips
SiLicoN AS SUBSTRATE FoR RF
Although silicon is an excellent material for the fabrication of high speed transistors and circuits, silicon as a substrate for RF circuits hardly couldn’t have been a worse choice. in an RF circuit all the components have a capacitive coupling to the resistive substrate. This means that at high frequencies part of the RF signal is dissipated in the substrate. This results in increased power consumption, and it makes it difficult to integrate passive components like striplines and inductors with a good quality factor. in return thermal noise generated by the resistive substrate is injected back into the circuit. Additionally, signals from one part of the circuit couple into different parts. This cross-talk makes it difficult to combine fast switching digital circuits with sensitive receiver circuits. For most of these problems designers and engineers have found a “workaround”. Multilevel metallization, trench isolation, shields and substrate contacts have been used to minimize undesired sub-strate interactions. However, the most effective solution still is to tackle the problem by its roots and to remove the silicon substrate altogether.
Figure 2: Silicon wafer transferred to a 10 µm thick polyimide foil.
Figure 3: The world’s thinnest fully integrated RFID tag circuit.
Maxwell11.2 March 2008
11
STEM cELLS
Stem cells are characterized by two properties: self-renewal and potency. Self-renewal means that the cells can go through numerous cycles of division while still remaining in the undifferentiated state. Potency is the capacity of stem cells to differentiate into specialized cell types. The differentiation process is not only “induced” by chemical triggers, but also depends on the structure of the surrounding tissues. Embryonic stem cells are found in the blastocyst, an early stage embryo. Embryonic stem cells are called totipotent or pluripo-tent because they can differentiate in any of the 200 different cell types of the human body. Human embryonic stem cell research is controversial because, with the present state of technology, it requires the destruction of a human embryo. Adult stem cells can be found in small quantities in most organs. Most adult stem cells are lineage restricted (multipotent or unipotent). Recently research groups in Japan and the US have succeeded in “reprogram-ming” fully differentiated cells from adults into pluripotent stem cells. Undoubtedly, these exciting results will give a strong impulse to this already fast developing field of research.
can better follow the intricate shapes of the human body. In a recently granted STW project for instance, a flexible flow- and-pressure sensor chip will be develo-ped at DIMES for application in 300 µm thick guide wires that are being used to diagnose the severity of a stenosis in bal-loon angioplasty procedures.
Stretchable circuits
The next step after flexible is obviously stretchable. The most appealing applica-tions of these flexible circuits can again be found in the medical domain. Stretchable chips will be able to comply with the twis-ting and stretching of skin, muscles and other tissues. Additionally, whereas flexi-ble ICs can be bent to follow cylindrical surfaces, stretchable circuits can be bent over spherical surfaces, which is impor-tant in for instance retinal implants.
Several approaches are being pursued to fabricate stretchable integrated circuits. One school of researchers considers the use of elastic substrates. Lacour et al. have found that very thin gold conductors on elastic PDMS membranes can survive an elongation of up to 100%. At ECTM -DIMES, Theodorous Zoumpoulidis and Marian Bartek follow a different approach to make stretchable ICs. Rather than using an elastic substrate, they partition the silicon into rigid islands that are con-nected by conductive springs (Fig. 4). The springs not only provide stretchability to the circuit, but also form the electrical interconnect between the rigid silicon is-lands containing the active elements.
Livingchips
From silicon processing to stem cell re-search seems an enormous step. Howe-ver, recent progress in stem cell research has demonstrated that it is about to mas-ter the three basic steps in IC-fabrication:
reproducable production by
differentia-tion of stem cells into specialised cells,
deposition of these cells on suitable
sub-strates, and cultivation of the cells into
designed patterns [3]. On the other hand
silicon technology is, as we have seen, adapting more and more biological fea-tures such as flexibility, stretchability and biocompatibility. These developments have brought both disciplines on an inter-section course.
At the present pace of research, it is very well conceivable that within a few years, the incorporation of living cells in micro-systems for diagnostic and therapeutic use becomes a practical reality. Similar to the growth or deposition and patterning of dielectrics and metals in today’s Micro-systems, in the future the flowcharts used for the fabrication of bio-MEMS may well specify the deposition of structured layers of (human) cells.
These cells may be used for sensing or ac-tuation. Already researchers from the Tok-yo University of Agriculture and Techno-logy in Japan have demonstrated how the periodic contraction of heart muscle cells can be used to generate useful amounts of electricity. Additionally, muscle cells may be used to actuate the movement of e.g. catheter tips or implants.
Concluding, progress in micro-electro-nics encompasses more than just cram-ming more and more components on a
silicon chip. Also in the field of substrate engineering significant progress is being made, extending the scope of silicon tech-nology to new fields of application. For information on positions for PhD’s, gra-duation projects and internships feel free to send an e-mail to Ronald Dekker via
r.dekker@tudelft.nl.
A
[1] www.intel.com/technology/ mooreslaw/index.html [2] www.tudelft.nl/live/pagina. jsp?id=44eb4562-80ca-4c4f-a625-98d05b2a67cd&lang=en [3] Science 317, 1366 (2007)Figure 4: Rigid silicon islands measuring 200•200 µm2 connected by metal springs
form a stretchable integrated circuit.
