Systems/Experiments

Nitrate sensor chips were tested for calibration curves (i.e. sensitivity, linearity, detection limit) and selectivity. The electrochemical measurements were conducted using a CH Instrument 660B potentiostat and a data-acquisition computer (Figure 1) Electrolyte (0.01 M NaOH) was freshly prepared with reagent-grade chemicals and triply-distilled DI water. For calibration, 0.1 M and 0.01 M nitrate standard were also prepared by dilution of freshly-made 1 M nitrate stock solution. Similarly, 1 M nitrite (NO_2-), fluoride (F-), chloride (Cl-), sulfate (SO₄^2-), phosphate (PO₄^3-), and potassium (K⁺) were prepared for selectivity test.

Figure 1: The bench-top experimental setup.

Problems Encountered

Sensor response is different from die to die on a wafer because of variance in electrode surface area. In order to obtain a meaningful sense on sensor performance, we tested 21 chips and analyzed their calibration curves.

Accomplishments

Two main accomplishments made during this research period are:

• Newly-designed nitrate sensor chips are tested for calibration curves and performance are characterized and tabulated
• Quantification of nitrate sensor selectivity to interfering ions in ground water

Calibration curves

The sensing chip designed the last year is incorporated with large counter electrode, polyimide insulation and electrode-protection layer, and microchannels. Figure 2 shows the SEM image of a sensing chip. It surpassed the initial design in terms of reliability, life-time, and reproducibility. However, sensor response is different from die to die even though they are from the same wafer. To estimate performance of sensors, we need to test a large number of chips to understand sensor parameters (i.e. detection limit, linear range, linearity, and sensitivity).

Figure 2: SEM image of the micromachined electrodes and microchannels.

 

A total of 21 chips were tested. Two chips were used to characterize silver oxide reference electrode. Eight chips are broken during experiment, mainly because polyimide protection layer was not fully adherent to silicon substrate and thus silver working electrode was protected badly. The rest of chips worked properly but three chips showed poor response (i.e. very-short linear range) because working electrode was biased to wrong potential by mistake. The calibration curves were obtained with the rest of good 8 chips. Figure 3 shows calibration of four chips, namely chip 55, 41, 86, and 76. Solid dark-gray line represents linear calibration before its slope begins changing. Light-gray dotted line represents the secondary calibration curve, which deviates from slope of the solid line. For example, chip 55 is linear from 0 to 2000 μ. From 2000 μ M to 10000 μ M, it is still linear but slopes changes. Table IV-1 summarizes calibration curves of all 8 chips. Slopes a₀and intercept a₁ are calculated with linear regression of charge data at each concentration. A calibration curve is expressed as y = a₀x + a₁, where y represents measured nitrate-reduction charge and x is concentration of nitrate standard. Sensitivity S equals 1/a₀ (C/ μ M) by definition. Linear range is where slope a₀ begins changing. Range varies from 500 μM to 2000 μM. It is encouraging result that our initial result was 1000 μM as reported the last year. Linearity (r²)is 0.99 in all cases except chip 11 and we note that calibration is sufficiently linear throughout the range. Detection limit is in strict sense defined as a critical concentration of analytes that is possible to be differentiated from blank solution with a certain confidence.  This is because analytic signal of blank solution has statistical distribution with mean X_b and standard deviation σ_b and if σ_b is large there is high probability that blank signals overlaps real nitrate signals. Detection limit is X_dl = X_b zσ_b in term of analytic signal (i.e. charge) and C_dl = X_dl / S in concentration unit, with z value of predetermined confidence level. However, it is not practical to obtain true σ_b because of limited sample size, we estimate it with sample standard deviation S_b and substitute z with t value of Student-t distribution (X_dl = X_b + tS_b). S_b was calculated with 5 background measurements, and thus z is substitute with t-value of 97.5% confidence and 4 degree of freedom (t=2.776). Detection limit also varies from die to die and it ranges from 4 to 75 μM.

Figure 3: Calibration curves of nitrate standard measured in sensor chip (a)~(c) with silver oxide reference electrode (d) with Ag/AgCl reference electrode

Chip 34 to 11 in Table 1 is tested with silver electrode anodized in electrolyte and it is less stable than commercial reference electrode. For chip 76 and 12, a commercial Ag/AgCl reference electrode was used for calibration. In comparison, the chip 76 and 12 has higher range of 5000 μM as in Figure IV-3 (d) and also life time is much longer than 6 chips with silver oxide reference electrode. This performance is what we can expect at most with current design with silver oxide reference electrode because of their unstable reference electrode. We concluded that: (1) linear range is from 500 to 2000 μM (2) linearity is 0.99 (3) detection limit is from 4 to 75 μM (4) reference electrode affects sensor performance in terms of range and lifetime. We reported polyurethane-coated wire reference electrode in the last literature, and it is important to incorporate it into the sensing chip to improve performance. 

 

Table  1: Performance of sensing chips

 

Selectivity of Nitrate Sensor chip

There are many kinds of inorganic compound in ground water such as cation, anion, neutrally solvated species, and trace metal (Table 2). They can potentially interfere with nitrate measurement as they usually increase analytic signal of sensor. Therefore, it is very important to quantify their interference. We select 6 different ionic species, nitrite (NO_2-), fluoride (F-), chloride (Cl-), sulfate (SO₄^2-), phosphate (PO₄^3-), and potassium (K⁺), and measured analytic signal with silver working electrode and interpret them as effective nitrate concentration.

 

Table 2: Ionic species in ground water

Figure 4.  Equivalent nitrate concentration of 6 ions of 100 μM and their mixture in 0.01M NaOH electrolyte.

At first, nitrate calibration curve are obtained. 1 M standard of 6 different ions are prepared and 100 μM each was measured with the same working electrode. Secondly, measured analytic signal are converted to equivalent nitrate concentration with the calibration curve. Their values are plotted in Figure IV-4. 100 μM nitrate response are depicted together for comparison. Except chloride, all ionic species has little effect on nitrate measurement. Even a mixed solution of all 6 ionic species (600 μM) has a little effect (2.78 μM in nitrate concentration). Therefore, we concluded that influence of six ions to nitrate measurement is negligible.

Future Directions

• Integration of Ag/AgCl Reference Electrode

PU-coated Ag/AgCl reference electrodes show reliable reference potential. We will work to integrate this electrode into our microsensor [May 2006 – July 2006].

• Long-term Qualification Test

Currently long-term qualification tests of nitrate sensor for field application are on-going. [Present-April 2006]

• Connecting to Wireless Network

The CENS-designed sensor interface board will be used to interface present sensor setup to wireless Motes. A graduate and a undergraduate student is working to make it fully functional. Data transmission and sensor control will be tested wirelessly in sensor network. [Present – September 2006]

• Nitrate Sensor Field Test

After bench-top performance tests are completed, nitrate sensor units will be deployed in the field to measure concentration of nitrate in real ground water. [September 2006- December 2006]

 

People

Jack W. Judy, Professor, UCLA
Ira Goldberg, Dr., UCLA
Dohyun Kim, GSR, UCLA