document

Abstract

This paper shows proppant induced pressure increase (i.e., “tip screenout”, “pack”, “body pack”, etc.) can relate to restricted vertical and lateral proppant distribution in hydraulic fractures. We discuss a way of interpreting the character of the pressure response during the proppant stages of a fracture stimulation. The pressure response during the proppant stages will determine the quality of the production response and the level of success that is achieved. The technology presented has been found to apply to all rock types and most formations.

This area of study is based upon common sense observations and extensive engineering. It fits the reservoir engineering of the producing character and advanced fracture treatment pressure interpretation on over 1,000 wells. When early proppant bridging occurs in the fracture, the lateral and vertical proppant distribution adjacent to the pay interval can be negatively affected. This results in less effective stimulation because the pay interval, both vertically and laterally, is not effectively propped. The fracture stimulation may have been “put away”; but the negative aspects of how the stimulation was designed and implemented may have a significant affect on the resultant production response.

To lay the basis for additional thought and investigation, there is discussion about deficiencies in overly simplified pre-treatment mini-fracture analysis procedures. These analysis methods are not focused on the proppant induced friction character, and therefore can not optimize fracture length. An evolved approach is presented, that addresses leakoff, fracture geometry, and proppant induced friction, to optimize the proppant distribution, and overall stimulation of the pay interval.

Introduction

The technology presented is derived from a significant volume of work that is based upon an integrated engineering approach to determine the effectiveness of the completion and stimulation method based on the production response. Over the last five years, our technical advancement has been extensive in the application of this area of study. The resultant production responses in many areas have been significant, supporting the credibility of the technology presented.

During the mid ‘90s, downward proppant movement (i.e., clustering, settling, convection1, 2, 3, 4) in hydraulic fractures received significant attention in technical papers, forums, meeting, etc. It was depicted as a dominant variable in hydraulic fracturing and, if not addressed, was the reason many wells did not properly produce. The proppant was theorized to move to the bottom of the fracture and not be adjacent to the pay interval. An approach for minimizing downward proppant movement was to “tip screenout” or “pack” the fracture.

In addition, the perspective existed for many years that a “tip screenout” is the ideal response. If conductivity is desired, then build the fracture conductivity by “screening out” or “packing the fracture”. This applies in high permeability, low modulus rocks; however, in low to moderate permeability rocks it may not apply. All fracture models have varying degrees of capability. How each of these models represents the distribution of proppant in the fracture is an evolving science.

During the early to mid ‘90s, many fractures were packed to minimize downward proppant movement. Through detailed reservoir engineering evaluation, these packed fractures were found to either have short effective fracture lengths or skin damage. Our initial evolution was to reduce the level of pressure pack during the proppant stages. Excessive levels of pack were obviously causing damage as a result of polymer dehydration, (“polymer squeeze”) on the formation and fracture. With proppant induced pressure increases of less than 1,000 psi, the reservoir engineering analysis continued to show effective fracture lengths shorter than expected. The design criteria evolved to a lower, < 500 psi, proppant induced pressure increase. The effective fracture length, from reservoir analysis, became longer, but still did not meet expectations. To improve the predictive capability of stimulation designs, customized fracture models were developed that matched the proppant induced friction character observed for various fluids, formations, fracture geometries, etc.

When evolving in this direction, it was found that mini-fracture evaluation and standard industry fracture model usage were flawed and did not work. Standard mini-frac evaluation was designed based on leakoff and fracture geometry assumptions during an era when a tip screenout was the “ideal” fracture treatment. These procedures are not tailored towards optimizing proppant distribution because they do not focus on the influence of proppant-induced friction.

During this same period, we began studying and observing the microseismic and tiltmeter imaging of hydraulic fractures5, 6 ^{and others}. There is a lot to be learned in this technical area. Specifically, how imaged fracture geometry relates to actual hydraulic dimensions, how hydraulic dimensions relate to propped dimensions, and finally how this leads into effective propped fracture dimensions? In our evaluation of fracture imaging projects, the one consistent observation was that proppant induced pressure increases retard lateral hydraulic fracture growth and promote height growth, usually upward.

We tied our interpretation of the fracture imaging projects with the reservoir engineering and fracture treatment net pressure interpretations of many wells. The lateral proppant distribution hypothesis was the consistent logic path that fit all scenarios. If proppant is entering the fracture and building proppant friction induced backpressure, then how is it efficiently distributing vertically and laterally adjacent to the pay interval? What affect does this have on the geometry?

If this was correct, the perspective on proppant induced pressure increases needed to evolve. A tip screenout may be screening out at a tip. What tip is it? Is it the hydraulic fracture tip or is it the proppant tip that has bridged off in the fracture? The hydraulic tip may be significantly further out in the formation than the tip that is screening out. Where is the tip that is packing? As Baree7 presented in 1991, an increased pressure at the fracture tip will cause fracture height growth.

This lateral proppant distribution hypothesis is an alternative explanation why fracture cleanup/polymer damage can not be proven to be the primary cause of short effective fracture lengths. The proppant may not be across the pay interval or heavily damaged in the packed region.

We began our evolution with this technical hypothesis during 1994. The success of optimizing proppant distribution was experienced quickly through improved production responses. Instead of immediately publishing results based upon a few case histories, it was decided to fully test the hypothesis and achieve long-term production data. This was an effort to ensure technical consistency and credibility. To date, the evolution has involved application on over 1,000 fracture treatments. It has been the primary element in the improved production responses achieved in many areas. We have found it has application in all rock types and formations.

The Proppant Distribution Hypothesis

Do proppant induced pressure increases relate to the production response that is realized? If a frac job is “put away”, independent of the pressure character, did we pump a good stimulation treatment that maximizes the production? Could the fracture treatment be designed and implemented differently to improve what occurs? How important is it to optimally, ideally, properly, etc. engineer a fracture stimulation? These are all questions we have struggled with for many years. This paper addresses these questions and shows it makes a big difference.

In many reservoirs, the fracture stimulation is the primary operation for optimizing the production response. If proppant induced pressure responses negatively affect the distribution of proppant in a fracture, then they definitely have a negative affect on the production response.

Figure 1 shows a comparison of three fracture treatment pressure responses. The pressure response without proppant is representative of what would happen if no proppant induced friction occurred. Because there is no proppant induced pressure affect, this would imply, if proppant was present, it is easily entering and distributing in the fracture.

The two pressure responses with proppant are typical of what can occur. In our investigation we asked ourselves if the production response was affected by the character of the proppant induced pressure response. Can a fracture treatment modeling and design process be developed to control the level of proppant friction that occurred? If so, will this maximize the production response? The answer is yes.

Figure 2 shows what is commonly called asymmetric proppant bridging. The proppant enters the fracture, the pressure inflects and the proppant continues bridging in the fracture. The rapid pressure decline after shut down shows that the proppant bridged asymmetrically within the fracture. With this pressure response, the proppant is not well distributed. A high proppant concentration exists for a short distance.

Figure 3 is termed, “Improved Proppant Distribution?”. The pressure still inflects upon proppant entry, but climbs less gradually. Intuitively, the proppant distribution with this response would be expected to be better than the response observed in Figure 2. However, the fracture imaging data shows fracture height growth with small proppant induced pressure increases. A look at the M-Site data tells a story that fits those specific conditions. The height growth started almost the same time the pressure inflected. The total magnitude of the pressure inflection may not be the critical variable. If small proppant induced pressure inflections negatively affect lateral proppant distribution, then Figure 3 may not be the desired response either. Further work is needed here.

The following examples will show, from a practical applications point of view, downward proppant movement via clustering, settling and convection is not a primary issue in hydraulic fracturing. The proppant-induced friction should be minimized to allow the proppant to distribute laterally and vertically in the fracture. Customized fracture models are required to account for the proppant-induced friction characteristics that exist for various treatment fluids, formations, fracture geometries, etc. The examples that follow will show the significant improvement in production response that can be realized.

Example 1: Deep Red Fork - Anadarko Basin

This project was implemented during 1996 and received an award in 1998 for Best Field Improvement Project in the Mid Continent Region8. All of the wells are offsets and are completed in the same consistent sand interval. Table No. 1 shows the pertinent reservoir properties.

(See Table 1 at the end of the article.)

There are three vintages of fracture treatments presented in Figures 4,5 and 6. The important variable to investigate is the difference in the treating pressure character.

Figure 4 is characteristic of a late 1970’s type of stimulation treatment which was typically 77,000 gallons of zirconate crosslinked fluid and 100,000 lbs of sand with 50,000 lbs of glass beads pumped at 15–20 bpm. Figure 4 shows injection rate, proppant concentration and net pressure. In this treatment, note the low proppant concentrations in ½ ppg increments, from 1 to 3 ppg. The net pressure does not build, as a result of proppant entering and moving down the fracture, suggesting the proppant should be easily distributing.

During the “convection days”, an interpretation of this fracture treatment might be that the proppant moved to below the pay interval because a packing response was not observed. There is no question that the proppant effectively distributed adjacent to this pay interval. Pressure buildup and production analyses indicate the existence of effective fracture lengths on the order of 250 ft, with finite conductivity. Modeling of the fracture treatment net pressure response suggests a 250-ft fracture half-length with low proppant concentrations. The fracture model and production analysis interpretations are consistent.

Figure 5 is the late 1980’s and early 1990’s fracture stimulation approach. Figure 5 also shows bottomhole injection rate, bottomhole proppant concentration, and net pressure. These treatments typically consisted of 167,000 gallons of crosslinked binary or CO₂ foam and 300,000 lbs of intermediate strength proppant pumped at 30 bpm. The net pressure inflects upward with proppant entry into the fracture. With the increases in proppant concentration, the rate of pressure climb increases, then the pressure breaks over flat. The net pressure increase, as a result of proppant induced effects, is 1,050 psi. This is not an untypical pressure response on many fracture treatments. In fact, in many areas, it is more the normal response than the exception. The question to ask is does it make a difference and does it affect the production response?

Pressure buildup and production analyses indicate a 250-ft infinite conductivity fracture. Modeling, of the fracture treatment net pressure response, suggests a 450-ft infinite conductivity fracture should exist. There is an obvious inconsistency between the fracture length predicted from the fracture pressure matching and what actually occurred. Figure 6 will show this inconsistency is a result of restricted lateral proppant distribution.

Figure 6 is one of the earlier treatments that began the testing of the proppant distribution hypothesis. These treatments typically consisted of 120,000 gallons of borate crosslinked fluid and 300,000 lbs of precured resin coated proppant pumped at 15–20 bpm. The character of the net pressure as proppant is entering and distributing within the fracture is significantly different than presented in Figure 5. It is more similar to the response observed with the well in Figure 4, which achieved good proppant distribution. Neither of these wells has the proppant bridging character shown in Figure 5. This reduced proppant effect was achieved with less fluid and injection rate. This shows the importance of coupling the complete process. The perforation and breakdown strategy were designed to control the fracture initiation character. This was linked with the pre-treatment diagnostic interpretation and design process.

If the proppant distribution hypothesis is correct, this treatment should result in a longer effective propped fracture length. The effective propped fracture length from pressure buildup and production analyses is 400 ft, while the predicted fracture length from net pressure modeling is on the order of 400 ft. By eliminating the asymmetric proppant bridging character, the proppant distribution is improved and the effective fracture length is longer.

Figures 7, 8, 9 are a comparison of the average production responses of the three different types of stimulation approaches. Figure 7 is a comparison of the average rate versus time and Figure 8 normalizes the production comparison for differences in reservoir and flowing pressures as well as reservoir thickness. On figure 8, observe that the early-time production is improved 3–4 fold and the late time production has stabilized at a 2–3 fold improvement.

Figure 9 shows a comparison of the average inverse productivity index versus time on a modified MDH scale for each of the three stimulation pressure responses. Using this type of analysis, Crafton9, 10 and Reitman11 have shown the slope of the curves relate directly to the reservoir transmissibility, (Kh/u), while the y intercept is a function of the effective wellbore radius or propped fracture length. The near parallel nature of the respective dp/q responses in Figure 9 suggests the average transmissibility for each group is similar. Also note that the extrapolated y intercept for the average dp/q for the wells with improved proppant distribution yields the longest effective propped fracture length. This interpretation suggests that the wells using the improved proppant distribution approach yielded better production results because of longer effective propped fracture lengths.

Example 2: Shallow Mesa Verde Oil - Wyoming

This project is in the shallow Mesa Verde, which produces oil from a relatively thick sand package. All of these wells are direct offsets and are completed in the same consistent sand interval. Table No. 2 below describes the pertinent reservoir properties.

(See Table 2 at the end of the article.)

There are two vintages of fracture treatments presented in Figures 10 and 11. Again, the important variable to scrutinize is the proppant induced pressure response.

Figure 10 is characteristic of the previous stimulation treatments which typically consisted of 15,000 gallons of gelled oil and 40,000 lbs of curable resin coated sand pumped at 35 bpm. Typically, 2–4 stages of this type were performed in each well. Figure 10 shows injection rate, proppant concentration and net pressure. When proppant enters the fracture, the pressure inflects and climbs. This is an indication that the proppant is not effectively distributing in the fracture. Production analysis indicated effective fracture lengths less than 50 ft with infinite conductivity. Modeling of the fracture treatment net pressure response suggested the fracture length should be on the order of 120 ft. This shows an inconsistency between the stimulation design and the fracture length that was achieved.

Figure 11 is typical of the new approach using 40–60,000 gallons of borate crosslinked fluid and 150,000 lbs of curable resin coated sand pumped at 30–40 bpm. The completion, using this stimulation design approach, is only one stage instead of the 2–4 stages previously implemented. By performing one stage, it resulted in significant cost saving to the operator. The net pressure does not inflect until later in the proppant stages and builds at a slower rate. The total pressure build is 150 psi as compared to the 450 psi shown in Figure 10.

The production analysis on the improved proppant distribution wells indicated a 150-ft effective fracture half-length. The modeling, of the fracture treatment net pressure response, suggested a fracture half-length on the order of 150 ft. Improving the proppant distribution has enhanced the effective fracture length and production response.

Figures 12, 13 show the benefit of implementing a proppant distribution design. Figure 12 shows a comparison of the average production rate versus time for five wells with the improved proppant distribution and six wells with the poor proppant distribution. The initial average production rate was increased 50%. After 35 months, the incremental recovery is 22 MBO per well.

Figure 13 is the inverse productivity index versus time in modified MDH format. This shows the transmissibility is similar between the two well groups. The wells with improved proppant distribution are shown to have longer effective fracture half-lengths.

Example 3: Generic hard rock example.

This example is typical of many moderate to high modulus reservoir rocks. This reservoir produces gas and has a history of developing complex fracture geometries. All of these wells are within close proximity. Table No. 3 describes the pertinent reservoir properties.

(See Table 3 at the end of the article.)

There are three types of fracture treatment approaches presented in Figures 14, 15, and 16. Again, scrutinize the differences in proppant induced pressure.

Figure 14 is characteristic of the previous stimulation treatment approach with 190,000 gallons of borate crosslinked fluid and 342,000 lbs of proppant pumped at 25 bpm. Figure 14 shows injection rate, proppant concentration and net pressure. The net pressure inflects when proppant enters the fracture and climbs an additional 3,600-psi before the end of the treatment. The fracture stimulation is “put away”. Does the character of the pressure response have an affect on the effective fracture length and production response? This is a high level of pressure increase and is considered an extreme pack.

A customized fracture model was developed that matched the proppant bridging character of the prior treatments. Using this model, the stimulation design was changed to reduce the proppant bridging character. Figure 15 shows the first evolution in techniques to reduce the proppant bridging and improve the distribution. This type of treatment consisted of about 250,000 gallons of borate crosslinked fluid and 340,000 lbs of proppant pumped at 40 bpm. Again, the net pressure inflects with the entry of proppant and climbs about 1,000 psi.

This reduced the proppant induced friction and showed a production improvement over prior treatments. The first well showed positive results, with the early-time production increase nearly four fold, from 500 Mcfpd to 2,000 Mcfpd.

Four additional fracture treatments were executed with this moderate level pack response. The customized fracture model was further refined and the next evolution was pursued to further minimize the proppant bridging.

Figure 16 shows the final evolution in approach used to improve the proppant distribution. These treatments typically consisted of 300,000 gallons of borate crosslinked fluid and 300,000 lbs of proppant pumped at 40 bpm.

In Figure 16, the net pressure does not inflect as a result of proppant entry and the maximum definable pack is a few 100 psi. The initial decline in net pressure is near-wellbore cleanup. The small pressure swings while pumping are a result of operational fluid QC problems. If the character of the pressure response during the proppant stages relates to the effectiveness of the stimulation, then the production results should be better than the previous two well types.

Figures 17, 18, 19, 20, 21, 22 show the significant production improvement achieved by improving the vertical and lateral proppant distribution. Based on vintage of completion, Figure 17 is the before and after average well rate comparison, while Figure 18 is the average well productivity comparison. A significant improvement in production response and productivity were observed, 300% early and 200% late time.

The new approach wells presented in Figures 17 and 18 are broken out based upon proppant induced packing response in Figures 19, 20, 21, 22. These figures breakout the wells based upon extreme packs, >1,500 psi, moderate packs, >500 psi, and light packs, < 500 psi. In Figure 19, the rate comparison of the average well response shows an improved production response with the fracture stimulations that demonstrated the least amount of proppant-induced friction. When using the normalized productivity comparison in Figure 20, the light packs produce almost twice as good as the moderate packs. Also in Figure 20, the light packs are significantly better than the extreme packs, 350% early and 200% late time.

Figure 21 is the modified MDH plot of inverse productivity index versus time. The slope of the lines are similar, indicating each group has comparable transmissibility. Figure 21 clearly shows, with each evolution in the proppant distribution technique, a longer fracture length is achieved. The oscillating character of the extreme packs and late-time quality of the moderate packs is associated with liquid loading effects.

Figure 22 is another way to evaluate the success of the proppant distribution technique. This illustration shows the vertical proppant distribution over the pay interval is also affected. Figure 22 shows leakoff coefficient obtained from the pre-treatment pump-in/shut-ins versus the production analysis derived flow capacity, kh. The leakoff character of low viscosity, water or linear gel, can be used to determine the formation permeability and productive capability of an interval.12

Figure 22 shows similar leakoff responses for most of the wells. This indicates comparable permeability and that the wells’ productive capability should be similar. However, there is a big difference in the flow capacity, kh, realized from the production analyses. The lower packed wells show a higher effective reservoir flow capacity, kh. This is a separate mechanism from the longer fracture half-lengths that are realized and supported by Figure 21. It can be inferred that reducing the proppant induced friction also improves the vertical proppant distribution adjacent to the pay interval. This allows more of the vertical reservoir thickness to contribute to production. A higher flow capacity, kh, is realized.

Discussion

This work is primarily focused on showing the impact of proppant-induced friction on the proppant distribution in a fracture and how this relates to the effective fracture length/production response. There are other variables and conclusions involved in this work that require additional investigation. We will summarize a few of these.

Polymer loading in the fracture as it relates to packing:

Packing a fracture is commonly known to cause polymer dehydration, which can lead to damage in the fracture and formation face. If an over-packed pressure response causes damage, what level of packing response will result in minimal damage? The ability of the formation to clean up the damage is a consideration. When the slurry stops moving in the fracture, high-pressure differentials can be placed on the slurry. This can cause polymer dehydration, thereby increasing the effective polymer loading in the fracture at job shut down.

This has implications on the design of the polymer loading. Under the premise presented, a low polymer loading is most effective if the proppant induced pressure differentials are minimized. If the lower viscosity of a low polymer loading causes more proppant-induced friction, it may not be the optimum fluid system to use. A slightly higher polymer loading without a pack may result in a lower effective polymer loading in the fracture at shut down. Assuming fracture closure does not occur quickly, the lower the polymer loading, the easier it is to break and the lower the viscosity during cleanup. This may promote improved fracture cleanup, as opposed to a packed, dehydrated, polymer system.

Mini-fracture methodology versus proppant distribution techniques:

Mini-fracture (datafrac, FET, etc.) analysis techniques are based upon leakoff and fracture dimensional growth assumptions. They are not focused upon proppant induced friction character. The differences between mini-frac methods and proppant distribution techniques are significant.

Proppant distribution designs are based upon an understanding of the actual proppant induced friction character that exists for the interaction of a given treatment fluid, formation, fracture geometry, etc. This requires the development of a customized model that matches what actually occurs for that combination of fluid and rock properties. Once the frictional character is effectively represented, it is possible to optimize a fracture stimulation design that incorporates the leakoff and fracture growth character with the proppant friction effects to maximize the vertical and lateral proppant distribution.

Packs, tip screenouts, etc. versus no packs:

Some fractures require some level of proppant induced pack for conductivity, to address embedment or near-well vertical communication due to limited perforation height. A small pack is also an effective way to control proppant flow back. A key component is the breaker design strategy, and ability of the formation to clean up any polymer damage.

A tip screenout is necessary in high permeability, low modulus rocks. The question still exists where is the propped fracture tip screening out relative to the hydraulic tip? There appears to be a significant difference between the propped and hydraulic fracture tips. This technology can be used to extend the realized propped fracture tip and achieve a longer, high conductivity fracture, with improved production response.

Fluid system selection.

The examples in this paper all use borate crosslinkers. This is not to be interpreted as a fluid of choice or superior fluid system for this application. The technology presented has been extensively applied with most fluid systems. The important consideration is that each fluid type has different rheologies and proppant induced friction character.

Conclusions

The character of the fracture treatment pressure response during the proppant stages determines the quality of stimulation that is achieved. Simply “putting the frac job away” may result in substantially lower production response than could occur. In low to moderate permeability formations, the perspective that a tip screenout is ideal, does not appear correct. Fracture treatments with less proppant-induced effects have longer effective fracture lengths and produce better.

Proppant induced pressure increases appear to restrict both vertical and lateral proppant distribution. This results in less effective proppant placement adjacent to the productive interval, which relates directly to less productivity.

Variations in treatment fluid rheology result in differences in the proppant-induced friction that occurs. Furthermore, each formation has differences in proppant induced friction character. It is possible to develop customized fracture models that represent the actual proppant induced friction for differences in fluid rheology, rock type, fracture complexity, etc. With this approach, the optimum stimulation can be performed that maximizes the proppant distribution in the fracture and maximizes production.

This investigation leads into questioning the use of too low of polymers. It also suggests that standard mini-frac approaches need to evolve to address the proppant induced friction character of different treatment fluids, formations, fracture geometries, etc.

Polymer loading in the fracture as it relates to packing:

Packing a fracture is commonly known to cause polymer dehydration, which can lead to damage in the fracture and formation face. If an over-packed pressure response causes damage, what level of packing response will result in minimal damage? The ability of the formation to clean up the damage is a consideration. When the slurry stops moving in the fracture, high-pressure differentials can be placed on the slurry. This can cause polymer dehydration, thereby increasing the effective polymer loading in the fracture at job shut down.

This has implications on the design of the polymer loading. Under the premise presented, a low polymer loading is most effective if the proppant induced pressure differentials are minimized. If the lower viscosity of a low polymer loading causes more proppant-induced friction, it may not be the optimum fluid system to use. A slightly higher polymer loading without a pack may result in a lower effective polymer loading in the fracture at shut down. Assuming fracture closure does not occur quickly, the lower the polymer loading, the easier it is to break and the lower the viscosity during cleanup. This may promote improved fracture cleanup, as opposed to a packed, dehydrated, polymer system.

Mini-fracture methodology versus proppant distribution techniques:

Mini-fracture (datafrac, FET, etc.) analysis techniques are based upon leakoff and fracture dimensional growth assumptions. They are not focused upon proppant induced friction character. The differences between mini-frac methods and proppant distribution techniques are significant.

Proppant distribution designs are based upon an understanding of the actual proppant induced friction character that exists for the interaction of a given treatment fluid, formation, fracture geometry, etc. This requires the development of a customized model that matches what actually occurs for that combination of fluid and rock properties. Once the frictional character is effectively represented, it is possible to optimize a fracture stimulation design that incorporates the leakoff and fracture growth character with the proppant friction effects to maximize the vertical and lateral proppant distribution.

Packs, tip screenouts, etc. versus no packs:

Some fractures require some level of proppant induced pack for conductivity, to address embedment or near-well vertical communication due to limited perforation height. A small pack is also an effective way to control proppant flow back. A key component is the breaker design strategy, and ability of the formation to clean up any polymer damage.

A tip screenout is necessary in high permeability, low modulus rocks. The question still exists where is the propped fracture tip screening out relative to the hydraulic tip? There appears to be a significant difference between the propped and hydraulic fracture tips. This technology can be used to extend the realized propped fracture tip and achieve a longer, high conductivity fracture, with improved production response.

Fluid system selection.

The examples in this paper all use borate crosslinkers. This is not to be interpreted as a fluid of choice or superior fluid system for this application. The technology presented has been extensively applied with most fluid systems. The important consideration is that each fluid type has different rheologies and proppant induced friction character.