UNIVERSITY OF WISCONSIN SYSTEM

SOLID WASTE RESEARCH PROGRAM

Student Project Report

Dissolved Phosphorus Removal using Steel Slag By-Products

May 2015

Student Investigator: Matt Tlachac

Advisor: Dr. Daniel Keymer

University of Wisconsin-Stevens Point

Introduction

The Lower Green Bay and Fox River Valley has been designated an Area of

Concern (AOC) in the Great Lakes Region. The watershed has recently come under an

Environmental Protection Agency (EPA) approved Total Maximum Daily Limit (TMDL) to

reduce discharges of Total Suspended Solids (TSS) and phosphorus. Over half of

these discharges are known to be coming from agricultural lands. Means to reduce

surface run-off sources of TSS and phosphorus are well understood and local

conservation agencies are in the process of working with producers to improve and

expand these practices. However, there is currently no means of containing dissolved

phosphorus from drain tile lines. Initial United States Department of Agriculture (USDA)

Agricultural Research Service work indicates that dissolved phosphorus concentrations

in tiled lands testing excessive for phosphorus and/or that have receive heavy

applications of manure applications are likely to have phosphorus concentrations in

discharge of 1 mg L-1 or greater (Joern et al., 1998). The AOC has a number of areas

identified as having high soil P concentrations and there are many large dairy

operations in the area. The need for these farm operations to land apply their manure

often leads to over saturation of phosphorus in agricultural fields.

Phosphorus in the environment binds readily to a handful of very specific

elements due to phosphorus’ most common molecular form in the environment:

phosphate. Phosphates consist of one phosphorus atom bonded to four oxygen atoms

and carry an inherent 3- charge; often seen as PO43-. The negative charge often repels

the phosphate molecule away from slightly negative soil particles. But clay particles in

particular often have positive trivalent elements at their center, allowing for a strong

attraction between the 3- charge of the phosphate and the 3+ charge of the clay particle

site. Iron and aluminum are the two most popular trivalent clay components that strongly
interact with phosphorus. Calcium and magnesium are also commonly associated with

having strong attractive forces with phosphorus.

Steel slag is an industrial by-product often used for beneficial uses under

Wisconsin Administrative Code NR 538; however there are still many thousands of tons

of slag being landfilled each year from operational foundries across the state. Facilities

that use electric arc furnaces and have more basic conditions produce slags that tend to

have higher phosphorus removal capabilities (Jones, 2015). Basic conditions refer to an

environment that have a higher concentration of OH- molecules than H+ atoms, which

occurs at pH values above 7. The more basic slag, often referred to as electric arc

furnace slag (EAF slag), will also provide a similar environment to many soils that have

basic parent materials such as limestone. Iron, aluminum, calcium, and magnesium are

often common constituents in the composition of EAF slag, which gives the slag its

natural tendency to sequester dissolved phosphorus. Being able to utilize EAF slag as

an adsorbent for phosphorus could provide multiple benefits: diverting slag waste from

landfills and limiting nonpoint phosphorus discharges to surface waters.

The objective of this research project was to quantify the effect of contact time on

phosphorus adsorption to two separate sources of electric arc furnace (EAF) slag. The

hypothesized outcome was if the contact time was increased between the phosphorus

solution and the EAF slag, then higher amounts of phosphorus would be removed from

the water. We also hypothesized that the slag source with the higher iron concentrations

would be the more effective filter media for removing phosphorus.

Methods

In order to quantify the effectiveness of phosphorus removal due to adsorption

onto EAF slag, a phosphorus solution with a known concentration the design of this

experiment was testing two different times of contact between and the EAF slag in a
laboratory setting. Two different sources of EAF slag were also tested side by side in

the experiment for removal effectiveness. The times of contact tested were 10 and 30

minutes, chosen to approximate peak and average tile drain flow rates, respectively. A

stock solution containing phosphorus was prepared to a 0.0645M concentration in the

lab by dissolving dibasic potassium phosphate (K2HPO4) in distilled water. The

phosphorus stock solution was then mixed into 400 liters of tap water to achieve an

influent solution with 1.2 mg L-1 phosphorus, contained in two 55 gallon HDPE drums

connected in parallel to allow equal draw from each drum. The phosphorus solution was

pumped up through the reaction vessels containing the EAF slag using a peristaltic

pump. Flow rate was estimated from the vessel volume and slag porosity (∑) using the

following equations:

Vpore = Vvessel * ∑

HRT = Vpore/Q

where HRT is the hydraulic residence time and Q represents flow rate. The vessels

ranged from 3.116 L to 2.987 L in volume. Each trial lasted for a duration of 48 hours,

and 10 total effluent samples were taken during each trial. The sample schedule was as

follows: 0, 0.5, 1, 1.5, 2, 3, 5, 9, 21, and 48 hours into each trial. The early phase of

each trial was sampled more intensively due to the findings of past studies showing

decreased removal rates of EAF slag over time. Each slag sample was run in triplicate

vessels for each contact time to ensure more reliable data, for a total of twelve trials.

Effluent samples were refrigerated at 4 degrees C for no more than two weeks, before

being filtered through a 0.45 µm filter. The filtrate from the test trials was run through a

Lachat QuikChem Autoanalyzer to measure the soluble reaction phosphorus (SRP)

concentration.
 The two slag sources varied in their particle size distribution. Slag source 1 from

Whitesville, IN, had been already crushed, whereas slag source 2, from Wisconsin

Rapids, had not been crushed yet. Manual size reduction of slag aggregates was used

to try and obtain a more homogenous mixture of particle sizes comprising the

adsorbent. In order to measure the distribution of particle sizes, a sieve analysis was

performed with mesh opening sizes as follows, from top to bottom, in millimeter

measurements: 12.7, 4.75, 2.36, 2.0, 1.18, and 1.0.

Performance metrics for the adsorption trials were calculated using formulas

derived by McGrath and Penn (2011). The maximum phosphorus (P) added calculation

determines what mass of phosphorus could be added to the slag filter before the

effluent phosphorus concentrations would rise above 99% of the original influent

concentration. Variables B and m were the same as explained in the Cumulative P

Removed (%) equation. The equation used to calculate it was:

𝐿𝐿𝐿𝐿 𝐵𝐵
𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑃𝑃 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 = −𝑚𝑚
mg P kg-1

Cumulative P removed (%) is a calculation that determines what percentage of

the maximum phosphorus loaded into the slag filters would be sequestered by the

media. The integrated expressions containing variables B and m in the equation below

are taken from curves fit for P removal with incremental P addition (Figures 1-4) and x is

the maximum P added (calculated above):

𝑥𝑥
∫0 (𝐵𝐵𝐵𝐵^𝑚𝑚𝑚𝑚)𝑑𝑑𝑑𝑑
𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑃𝑃 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 (%) = 𝑥𝑥

mg P Cumulative P removed is also expressed as a mass value with units of mg/kg.

The equation does utilize a couple of the previously mentioned calculations. The

equation used was:

 (𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑃𝑃 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 (%)
𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑃𝑃 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 = 𝑀𝑀𝑀𝑀𝑀𝑀 𝑃𝑃 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 ∗ �100)

The lifespan of each filter was calculated to determine how many days the filter

could last until all the phosphorus sequestering ability was spent. The equation used

was:

𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼 [𝑃𝑃]∗𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅

𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿 = 𝑀𝑀𝑀𝑀𝑀𝑀 𝑃𝑃 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 ∗ 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑜𝑜𝑜𝑜 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆

Results

EAF Slag 1 had a more homogenous composition of aggregates according to the

particle size distribution experiment. 78.61% of the material was between 12.7 and 4.75

mm in diameter. EAF Slag 2 had a higher diversity of particle sizes with 28.41%,

58.38%, and 9.89% in the 12.7+, 12.7-4.75, and 4.75-2.36 mm classes respectively.

Slag 2 was expected to have more diversity in its particle sizes due to the manual

crushing that was needed in order to obtain the more homogenous aggregate size.

However, Slag 1 had a higher proportion of aggregates smaller than 4.75 mm (20.3%

versus 13.2% for Slag 2). The results for the particle size distribution can be found in

Table 1 for both EAF slags.

The chemical composition of EAF Slag 1 revealed a high concentration of

phosphorus binding elements that were expected to be present. Iron and calcium were

the two largest components with 23.79% and 22.03% respectively. The next three

largest components were magnesium, manganese, and aluminum; with 6.04%, 4.09%

and 2.29%, respectively. Table 2 gives a complete chemical composition of Slag 1; a

composition of Slag 2 was not available at the time of the study.

The flow rate was derived from a calculation utilizing the porosity of the slag in

use. With an average porosity measurement of 0.467, the average pore space in the
reaction vessel came out to be 1.425 L. When the goal was to have a 10 minute

retention time, calculations told us we wanted a flow of 2.375 ml/s. When our goal was

to have a 30 min retention time, the flow calculation came to 0.792 ml/s.

The maximum P added calculation told us that EAF Slag 1 with a 30-minute

contact time would be able to handle the highest phosphorus loading rate. The

treatment condition that would accept the lowest phosphorus loading rate would be EAF

Slag 1 with a 10-minute contact time. The max P added calculation for EAF Slag 2 with

a 10-minute contact time seems to have been skewed by some measure within the

calculation. After comparing the rest of the data, Slag 2 with a 10-minute contact time

would have been expected to have the lowest allowable phosphorus loading rate.

The lifespan calculations for the different treatment conditions varied quite a bit.

The longest lifespan of 44 days belonged to Slag 1 with a 30-minute contact time, and

the shortest lifespan of 2 days belong to Slag 1 with a 10-minute contact time. Slag 2’s

lower readiness to remove phosphorus would explain why the filters would have a

longer lifespan. The needed mass of phosphorus to saturate every bonding site would

require a longer time to reach with Slag 2.

Within the same contact time treatment, EAF Slag 1 had a higher Cumulative P

Removal (%) in the 10-minute trials. Slag 1 removed approximately 20% of the

phosphorus on average, and Slag 2 removed approximately 4% consistently. The 30-

minute contact time treatment did not produce significant differences in percent

phosphorus removed between the two slag sources.

The mass of phosphorus removed per mass of slag had the highest result with

EAF Slag 1 with a 30-minute contact time with 117 mg P kg-1. The lowest removal rate

was seen using Slag 2 with a 30-minute contact time with a value of 9 mg P kg-1.
 Even though replicate trials of each treatment condition were carried out, similar

results were rarely seen. Attempting to hold all variables constant proved to be more

difficult than thought. Water behavior and flow inside the reaction vessel also could

have varied from trial to trial, which could explain the variability witnessed in the

calculations. In future studies, more replicate trials will need to be completed in order to

give a more concise data set.

Discussion

Slag 1 was more than 3 times higher than Slag 2 in the 10-minute contact time

treatment, and also higher in the 30-minute contact time treatment. This conclusion also

applies to the cumulative mass of phosphorus removed. The particle size distribution of

the two EAF slags would support this conclusion as well due to Slag 1’s higher portion

of smaller size aggregates. The prevalence of those smaller size aggregates means

there is more surface area within the EAF slag media for the phosphorus to interact and

bind to those phosphorus attracting elements such as: calcium, magnesium, aluminum,

and iron. C. J. Penn et al. also found the concentrations of those 4 elements in

particular to be strong indicators of a filters phosphorus removal capacity (McGrath and

Penn, 2011) Table 2 confirms that those 4 elements are present in relatively high

concentrations within that slag.

The impact of contact time has two different stories when comparing the data

within each slag source. For Slag 1, the longer contact time had created a difference in

the calculated max P added, lifespan of the filters, and the cumulative mass of P

removed. The max P added values for the longer contact time suggest those trials could

have accepted over twice as much phosphorus loading than the shorter trial and still

produce a clean effluent. Finding that contact time is a significant factor is consistent

with what C. J. Penn found in his 2011 study of EAF slag flow through phosphorus
removal filters (McGrath and Penn, 2011). Slag 2 on the other hand did not exhibit such

predictable results. Although the removal percentage went up, many of the other

calculated factors do not follow the same trend. Even within the same treatment,

different trials produced varied results.

As with any project, a few limitations stood out as the trials were carried out and

unfolded. The lack of constant mixing in the 55-gallon drums could have led to uneven

loading rates of the dissolved phosphorus to the filters. With slag 2 not being pretreated

for homogenous aggregate size, manual crushing of the aggregates was required. I

believe this fact heavily influenced the difference in the particle size distribution

witnessed between the two slags. While running the trials with 30-minute contact times,

periodic observation of the discharge flows would lead one to believe that steady even

flows was not being achieved. The construction and design of the equipment was often

correct on paper, but would run into issues in reality. In order to avoid these issues,

running each trial separately; one bin at a time would be advised.

Future studies in the area of dissolved phosphorus removal using EAF slag

adsorption could research the ease and availability of phosphorus reclamation out of the

slag filters. If possible, the reclaimed phosphorus and nutrients could then be reapplied

to the field it came from, or a surrounding area in need of phosphorus, but be applied in

a more conservation minded method. Running multiple filters side by side, with some in

service, and some out of service in “reclaim” stages could be a potential layout. From

the data generated by this laboratory study, a pilot scale project should be conducted in

order to further investigate the feasibility of using EAF slag as a phosphorus removal

structure for agricultural tile drainage fluid.

 Tables and Figures

Table 1. Particle size distribution for both EAF slags 1 and 2.

Slag 1 Slag 2
% %
Particle Sizes (mm) weight Particle Sizes (mm) weight
12.7 + 0.00 12.7 + 28.41
12.7-4.75 79.61 12.7-4.75 58.38
4.75-2.36 16.58 4.75-2.36 9.89
2.36-2 0.81 2.36-2 1.19
2-1.18 1.31 2-1.18 1.38
1.18-1 0.20 1.18-1 0.15
<1 1.36 <1 0.59
Unaccounted 0.13 Unaccounted 0.01
Total 100.00 Total 100.00

Table 2. Elemental composition of EAF slag 1. Results were provided by Whitesville Mill Services. Metal and
trace element concentrations were found using a inductively coupled plasma optical emission spectrometer.

Material
Slag 1
Element Concentration (PPM)
Aluminum 22869.00
Antimony <0.278
Aresenic <0.036
Barium 395.00
Beryllium 0.80
Cadmium 3.33
Calcium 220272.00
Chromium 3200.00
Cobalt 6.51
Copper 66.50
Iron 237885.00
Lead 8.89
Magnesium 60375.00
Manganese 40941.00
Molybdenum 13.50
Nickel 5.01
Potassium 51.40
Selenium <0.037
Silver <0.052
Sodium 237.00
Sulfur 2495.00
 Thallium 17.10
Titanium 1502.00
Vanadium 641.00
Zinc 41.40

Table 3 shows various calculations for the trials with 10 minute residence times. The Maximum P Added
calculation determines how much P could have been added to each slag filter before effluent P concentrations
rose about 1% of the original concentration. The lifespan calculation determines how many days the slag
filter could remain effective before becoming “spent”. Cumulative P Removal % is a calculation to determine
what percentage of the P added to the slag was sequestered by the filter. The mass of P removed during each
trial is quantified by the Cumulative P Removed calculation.

10 Minute Contact Time

Slag 1 Slag 2
Trial 1 2 3 4 5 6
Maximum P Added (mg) 117.831 106.934 157.727 499.581 377.568 353.813
Lifespan (days) 2.570 2.608 3.777 7.712 6.607 5.970
Cumulative P Removal (%) 17.715 23.275 20.975 4.467 4.940 5.636
Cumulative P Removed (mg/kg) 20.874 24.889 33.083 22.316 18.652 19.941

Table 4 shows various calculations for the trials with 30 minute residence times. The Maximum P Added
calculation determines how much P could have been added to each slag filter before effluent P concentrations
rose about 1% of the original concentration. The lifespan calculation determines how many days the slag
filter could remain effective before becoming “spent”. Cumulative P Removal % is a calculation to determine
what percentage of the P added to the slag was sequestered by the filter. The mass of P removed during each
trial is quantified by the Cumulative P Removed calculation.

30 Minute Contact Time

Slag 1 Slag 2
Trial 1 2 3 4 5 6
Maximum P Added (mg) 632.162 493.400 441.344 201.621 75.017 69.690
Lifespan (days) 44.436 33.265 30.292 12.613 4.593 4.182
Cumulative P Removal (%) 18.649 18.877 18.478 16.059 15.651 13.118
Cumulative P Removed (mg/kg) 117.892 93.139 81.552 32.378 11.741 9.142
 P removed (%)

y = 96.94e-0.029x
R² = 0.9996
y = 78.237e-0.037x
R² = 0.7458
y = 110.51e-0.044x
P added (mg/kg) R² = 0.9794

Figure 1 shows slag 1’s ability to remove phosphorus when 10 minutes of contact is allowed. Trial A is
depicted with diamonds, has a trendline equation of Y=78.237e-0.037x and an R2 value of 0.7458. Trial B is
depicted with squares, has a trendline equation of Y=110.51e-.044x and has an R2 value of 0.9794. Trial C is
depicted with triangles, has a trendline equation of Y=96.94e-0.029x and has an R2 value of 0.9996.
P removed (%)

y = y14.055e
= 16.954e -0.008x
-0.007x
y = 12.157e-0.005x
R² =R²0.8659
= 0.7854R² = 0.6731

P added (mg/kg)

Figure 2 shows slag 2’s ability to remove phosphorus when 10 minutes of contact is allowed. Trial D is
depicted with diamonds, has a trendline equation of Y=12.157e-0.005x and an R2 value of 0.6731. Trial E is
depicted with squares, has a trendline equation of Y=14.055e-0.007x and an R2 value of 0.8659. Trial F is
depicted with triangles, has a trendline equation of Y=16.954e-0.008x and has an R2 value of 0.7854.
 y = 83.524e-0.007x
y = 0.5234
R² 84.826e-0.01x
-0.009x
y ==82.553e
R² = 0.8921
R² = 0.4618
P removed (%)

P added (mg/kg)

Figure 3 shows slag 1’s ability to remove phosphorus when 30 minutes of contact is allowed. Trial A is
depicted with diamonds, has a trendline equation of Y=83.524e-0.007x and an R2 value of 0.5234. Trial B is
depicted with squares, has a trendline equation of Y=84.826e-0.009x and has an R2 value of 0.8921. Trial C is
depicted with triangles, has a trendline equation of Y=82.553e-0.01x and has an R2 value of 0.4618.

y = 68.995e-0.021x
P removed (%)

R² = 0.1024

y = 66.749e-0.056x
R² = 0.8835
y = 53.107e-0.057x
R² = 0.8352

P added (mg/kg)

Figure 4 shows slag 2’s ability to remove phosphorus when 30 minutes of contact is allowed. Trial D is
depicted with diamonds, has a trendline equation of Y=68.995e-0.021x and an R2 value of 0.1024. Trial E is
depicted with squares, has a trendline equation of Y=66.749e-0.056x and an R2 value of 0.8835. Trial F is
depicted with triangles, has a trendline equation of Y=53.107e-0.057x and has an R2 value of 0.8352.
 References

Fox, Garey, Derek Heeren, Joshua M. McGarth, Chad J. Penn, Elliot Rounds.

2012. Journal of Environmental Quality. Trapping Phosphorus in Runoff with a

Phosphorus Removal Structure. Vol 41: 672-679.

Joern, B. C., R. R. Simard, J. T. Sims. 1998. Journal of Environmental Quality.

Phosphorus Loss in Agricultural Drainage: Historical Perspective and Current Research.

Vol 27: 277-293.

Jones, Jeremy A. T. 2015. American Iron and Steel Institute. Electric Arc

Furnace Steelmaking. Webpage.

https://www.steel.org/Making%20Steel/How%20Its%20Made/Processes/Processes%20

Info/Electric%20Arc%20Furnace%20Steelmaking.aspx

McGrath, Joshua. M., Chad J. Penn. 2011. Journal of Water Resource and

Protection. Predicting Phosphorus Sorption onto Steel Slag Using a Flow-through

Approach with Application to a Pilot Scale System. Vol 3: 235-244.