Flourine Improves Polysilicon Resistor Stability

Brian Dance, Contributing Editor -- Semiconductor International, 8/1/2001

A collaborative study has shown that implanting fluorine can benefit the long-term stability of boron-doped polysilicon resistors for ICs. The fluorine prevents the access of hydrogen to dangling chemical bonds, breaking silicon-hydrogen bonds and recreating grain boundary carrier traps.

This long-term stability is important because polysilicon resistors are used in demanding applications such as balanced analog circuits, in which a small uncontrolled change in resistance values can prevent a circuit from working satisfactorily. This work has been done by Matts Rydberg of the Ångström Laboratory at the University of Uppsala (Sweden) and Ulf Smith of Ericsson Microelectronics AB (Kista, Sweden).

As reported in the August issue of Materials Science in Semiconductor Processing , the researchers deposited a 550 nm-thick polysilicon film on a 900 nm-thick silicon dioxide film with a low-pressure chemical vapor deposition (LPCVD) system at 630°C. The oxide layer had been thermally grown on a 20 W.cm p-type <100> silicon wafer. After the polysilicon was deposited, a 550 nm-thick SiO₂ layer was deposited at 400°C. The grain structure of the polysilicon was set by annealing, and the surface oxide was etched away to prepare samples for doping.

The group implanted the polysilicon films with 80 keV boron ions, at doses of 5 × 10¹⁴ and 1 × 10¹⁵ cm^-2. A 650 nm cap of oxide was deposited at 400°C before the samples were annealed in nitrogen atmosphere for 30 min at 1000°C. This oxide was etched away before implantation of ¹⁹F ions at 120 keV, an energy that placed the fluorine near the center of the polysilicon film. The ion doses ranged up to 5 × 10¹⁵ cm^-2. A 650 nm-thick oxide cap was deposited before the samples were annealed in nitrogen at 750°C for 30 min. This fluorine annealing temperature was a compromise between the adequate removal of implantation damage and minimal out-diffusion of fluorine atoms.

Resistors measuring 340 × 20 µm were defined by photolithography. Good ohmic aluminum alloy contacts were fabricated by alloying at 450°C for 20 min in a hydrogen and nitrogen ambient. The samples were nitride passivated by the PECVD deposition of 900 nm of SiN before hermetic sealing.

The researchers made various measurements to evaluate fluorine doping effects. Polysilicon film thickness was measured by SEM; TEM gave data on the polysilicon grain size; Hall measurement provided details of the active carrier concentrations and effective mobilities; and stress tests were performed at 100 and 150°C for 2200 hr to study resistor stability improvements. At the current densities of 3 × 10⁴ A.cm^-2, the resistor temperatures rose to 115 and 165°C.

Cross-section TEMs showed grains extended from the oxide interface up to the polysilicon film surface, showing that a 750°C anneal of the fluorine was adequate to recrystallize the polysilicon film. The measured grain size was 150 ±30 nm. Excess resistivity was used as a measure of the removal of fluorine implantation damage during the anneal, excess resistivity being taken as resistivity over that of a sample that had been annealed at 1000°C. For temperatures above 750°C, implantation damage was reduced to its asymptotic value at 750°C. The graph fits with an Arrhenius plot with a low activation energy of 1.4 ±0.1 eV, compared with polysilicon recrystallization's activation energy of 2.5 eV. Secondary ion mass spectrometry (SIMS) showed that little fluorine was lost during annealing at 750°C. A significant increase in Hall mobility was found when fluorine was added, in agreement with a model proposed by the researchers.

A gradual increase in resistivity under thermal and electrical stress was measured over 2200 hr, attributed to a gradual removal of hydrogen from dangling bonds in grain boundaries. The drift depends on the amount of hydrogen that blocks dangling bonds, so at least some fluorine accessed the dangling bond sites, where it blocked access for hydrogen atoms and remained attached during the stress test. As the amount of fluorine added to the polysilicon increased, the number of sites blocked by it rose and the resistivity drift became smaller. The researchers say the beneficial effect of fluorine saturates at a drift rate that is half that of boron-doped films without fluorine. The trap density in fluoride-doped films is comparable to that in films doped only with boron.